Then on Friday, the Attorney General admitted to what may turn out an even bigger SFO blunder. It had erroneously circulated privileged defence documents to the prosecution in the Nadir trial. Suddenly, accusations that it has been running a plot to undermine Asil Nadir's defence began to look almost plausible.
Failure at the SFO has become a depressingly familiar story. By now it is probably not going too far to say that the organisation has come to be regarded among the City and the legal establishment as something of a joke. Its inability to obtain convictions or, when it does, serious punishments, is notorious. Its credibility is severely compromised.
As Alistair Darling, Labour's spokesman on home affairs, put it last week: 'The Government must recognise that the public are losing confidence in the system . . . We must have a new centralised body - not the SFO.'
Most worrying of all, the SFO still has the mother of all fraud trials to come - the Maxwell affair. This is already straining its resources to breaking point. On its past record, the question must be, can the SFO cope, is it up to the job?
And as the allegations and revelations pile up in the Nadir prosecution, another question is beginning to be asked: is there something rotten about the SFO, have its officers got out of control?
A few days before the collapse of the Duralite case, the SFO scored one of the most embarrassing own-goals of the year with its failure to get Roger Levitt, accused of fraud and the theft of millions of pounds, sent to jail. Levitt emerged from the court with a grin a mile wide and an expression that said 'incredible, isn't it?'. To the man in the street, it was indeed incredible that, after all its efforts, the SFO accepted his plea to a minor offence. The manner of the fraud office's failure makes a remarkable and revealing tale.
The Levitt Group, an investment management firm, collapsed three years ago and the SFO was called in to investigate allegations of fraud. As so often in the past, it built up a hugely complex and cumbersome case of 60 charges involving the misappropriation of pounds 21m. The case was whittled down to 21 counts against Levitt presented in pre-trial hearings last February. The judge decided this was still too much and ordered them to be split into two trials. The SFO, under the leadership of George Staple, a corporate lawyer, then made what many criminal lawyers consider a basic misjudgement. It opted to bring a charge of 'fraudulent trading' against Levitt in the first trial.
'Why didn't they just charge him with theft?' asks one lawyer who has faced the SFO in a different fraud trial. 'Either Levitt did steal or he didn't, and any jury can understand that. Fraudulent trading, however, is a notoriously difficult charge to get through.'
The SFO assumed that this charge would include all Levitt's alleged misdemeanours, such as lying to bankers to obtain loans and misappropriating clients' funds. But Levitt's defence argued that much of the evidence relating to these allegations had nothing to do with the specific charge of fraudulent trading. At another pre-trial hearing in May, the judge accepted this argument. The SFO suddenly found half its most damning evidence excluded from the trial.
Having been outmanoeuvred, it could have appealed against this decision but confidently decided not to. It felt its case was still safe, partly because the judge had ruled that both sides should refer during the trial to Levitt's procurement of the pounds 21m in purely neutral terms.
But when it came to the trial, Jonathan Goldberg QC, Levitt's counsel, pulled a fast one. He declared in his opening speech that Levitt clearly had faith that his company was not going bust because he pumped in pounds 21m of his own money, 'honestly come by', to keep it afloat. The implication was that Levitt could not have been involved in fraud.
Because it had lost the earlier argument over evidence, however, the SFO was not allowed to allege that nearly all the money had actually been raised by Levitt under false pretences. The prosecution had been outmanoeuvred again.
In growing panic, the SFO now asked for the jury to be dismissed and the trial to start all over again. It claimed that the defence had gone further than it should have in claiming the pounds 21m was Levitt's own money. But although Mr Justice Laws rebuked Mr Goldberg for overstepping the line, he refused a retrial.
In a legal contest as uncertain as a fraud trial, the SFO was now staring possible defeat in the face. Realising the prosecution was on the ropes, Levitt then offered a deal. A week earlier, Terry Ramsden, a playboy and alleged fraudster, had got off all charges against him by accepting four relatively insignificant counts of 'recklessness'. He received a suspended sentence.
Spotting a good opportunity, Levitt offered the SFO a similar deal. He would accept the relatively trivial charge of misleading Fimbra, the investment watchdog, if all other charges of fraud and theft against him were dropped.
Although common in the US, plea bargaining does not technically exist in this country. By a typically British system of nods and winks, however, plea bargains take place none the less - and Mr Staple is an enthusiastic supporter of this procedure. Levitt's terms were that he would offer the plea only if it meant he would not have to go to jail.
Mr Staple and David Cocks QC, the SFO's counsel, were now so desperate to get any result against Levitt that they accepted the terms. It took Mr Justice Laws about half an hour to decide that Levitt's plea would not attract a custodial sentence - the misdemeanour he was admitting to was simply not that serious.
So the deal was done. Outmanoeuvred all the way down the line, the SFO got a result of some sort that it will no doubt claim as a success. The truth, however, is that Levitt skipped out of the Old Bailey a free man, with only 180 hours of community service to do instead of the years in prison he might have got had the original charges stuck. And the SFO's reputation as an effective bulwark against City fraud slipped another few notches.
The SFO was set up in the late 1980s to provide a more efficient and effective means of combating big City crime than the ordinary Fraud Squad or the Department of Trade and Industry. The idea was to combine lawyers, policemen and accountants in a single body that would both investigate and prosecute fraud cases.
It has scored some notable successes. The jailing of Ernest Saunders and several other Guinness defendants was the first bull's-eye, although in this case most of the investigation had been done by DTI inspectors, leaving the SFO to present the case in court. It also secured the conviction of Peter Clowes for stealing pounds 100m from his investment clients.
The SFO itself claims a high success rate. Since 1988, it has prosecuted 270 defendants and obtained 169 convictions. However, 100 of these were guilty pleas; in cases where defendants have pleaded not guilty, therefore, the SFO has won just 69 convictions. It has failed in 101 cases.
Why these failures keep happening is not easy to analyse because the SFO is secretive about the way it works and Mr Staple refuses, except on very rare occasions, to talk to outsiders. There is no clear pattern to what goes wrong with SFO cases - each time, the reasons seem to be different.
'One trouble with the SFO', says a lawyer with long experience of defending fraud cases, 'is that it tries to combine investigators and prosecutors. We're not used to that system in this country, and it doesn't work. The prosecutors - the lawyers and accountants - get too involved in the investigations. They can't see the wood for the trees.'
SFO investigations are certainly notorious for taking a long time. The Levitt trial, for example, took three years to bring to court. SFO investigators, like the DTI, have draconian powers under the Criminal Justice Act 1987 to oblige people to give evidence on pain of imprisonment. According to one defence lawyer, 'they have a nasty habit of sending a fax to someone demanding their immediate compliance, which tends to make people's eyes water.
'The delay in getting cases to trial is partly caused by defendants using every wrinkle the legal system allows to slow things down, as Saunders himself did in countless pre-trial hearings before the Guinness trial began.'
But SFO investigations still drag on for months and years, to the bewilderment of many lawyers. In one case, says a lawyer, 'the SFO took three years to investigate, but at the trial I didn't see any evidence I had not known about at the beginning'.
Yet investigation, however slow, is not the SFO's weakest suit. Where it seems to fall short again and again is in prosecution - constructing effective cases that will get a conviction.
It should have learnt a valuable lesson from the Guinness trials. It began by charging the defendants with just about every offence it could think of, but this was far too confusing for the legal system and for any jury. As the prosecution case was simplified and reduced, however, it became stronger, finally resulting in convictions.
Yet several years later, the SFO was still trying to launch its prosecution of Levitt with no fewer than 60 charges and having to split its cumbersome case into two trials.
Meanwhile, the Blue Arrow prosecutions against two City firms and several individuals were so complex and technical that they should probably never have been brought in the first place.
Inevitably, the cases all collapsed.
In other cases, the SFO seems to have swung too far the other way. In its case against Thomas Ward, the final Guinness defendant, the SFO made a narrow charge of the theft of pounds 5.2m. It then found its case was too narrow to present vital evidence of Mr Ward's deep involvement in the Guinness share support operation. As a result, he was acquitted.
What these cases all seem to demonstrate is a severe deficiency of legal judgement at the SFO. Time and again, it chooses the wrong tactics when presenting a case, and allows the opposition to run rings around it. 'There appears to be no clarity of thought there,' says a defence lawyer. 'Anyone can identify all the offences revealed in evidence. But it takes more skill to construct a case you can actually win.'
Why the SFO should suffer from this weakness is less clear. Some lawyers believe it should be run by an experienced criminal lawyer rather than a civil lawyer like Mr Staple. Others believe that too many SFO lawyers are second-rate - bureaucrats rather than litigators.
Certainly, doubts about the quality of the SFO's staff are continually reinforced by its repeated blunders.
Last week, for example, the Duralite case collapsed after what appears to have been straightforward SFO blundering. Three men - Malcolm Johnson, Mungo Park and Alex Dann - were charged with fraud and theft in connection with the sale of shares in an allegedly phoney US company. But the SFO was found to have passed confidential information to Clifford Chance, the City legal firm, which was pursuing a separate civil action against Mr Johnson. (The SFO refuses to comment, even though Mr Staple was head of litigation at Clifford Chance before he moved to the SFO.) The Old Bailey judge ruled that the three defendants had no case to answer.
On another occasion last year, an SFO official was jailed for trying to sell confidential documents.
But it is over the Polly Peck affair that many observers have begun to worry that the SFO may have gone completely off the rails. Earlier this year, an SFO lawyer was accused of forging a letter purporting to come from Sir David Steel, former leader of the Liberal Democrats. When Mr Steel objected, he was given the feeble excuse that the letter was an 'April fool'. Many observers began to wonder if some SFO officials were getting out of control.
This was relatively minor, however. More shocking was the SFO's own accusation that Asil Nadir had tried to bribe Mr Justice Tucker, the trial judge, in order to get possession of his passport to flee the country.
The judge himself was clearly as astonished as everyone else. The SFO, however, failed to produce a shred of evidence to back up its accusation.
''No one else would have had the arrogance to suggest such a thing,' comments one bewildered lawyer.
Still more embarrassing for the SFO, its counsel, Robert Owen, had told the judge he was likely to be interviewed by the police. This was refuted when, several weeks later, the investigating officer stated in court that he had never made any request to interview the judge.
Later still, two people turned up in Cyprus claiming that they had been put up to making the bribery allegations against Nadir and the judge by the SFO itself. The real truth of the matter may never be known, but the SFO's reputation is sufficiently tarnished for doubts to remain as to its conduct.
And there was more to come. Last summer, Michael Mates, the then Northern Ireland Secretary and a strong Nadir supporter, accused the SFO of plotting to destabilise Nadir's defence, and demanded an inquiry. Although fiercely denied by the Government and the SFO at the time, his allegations received dramatic support last Friday when the Attorney General admitted that privileged documents for Mr Nadir's defence had been erroneously circulated by the SFO to the prosecution.
Even if this was merely a mistake, it was still another shot in the foot by the SFO, a blunder the Fraud Office can ill afford at the moment. Already, Mr Nadir's supporters are suggesting that the Polly Peck prosecution should be abandoned because of the error. And the SFO's reputation has sunk still further.
There is little overt support in government circles for modifying or closing down the SFO at this stage.
Indeed, last Thursday Sir Nicholas Lyell, the Attorney General, defended the SFO's handling of the Levitt case despite the unsatisfactory outcome. He told Parliament that the prosecutions had been fairly and vigorously conducted.
But that is not enough to repair the SFO's damaged image with the public. If it screws up the Maxwell case next year, it may as well start writing its own obituary.