The inexperienced train driver Michael Hodder may have "misinterpreted" rather than simply missed a red signal moments before last October's Paddington rail crash, says a third interim report into the accident.
The signals on the gantry carrying the key signal 109, which Mr Hodder, 31, went through, were of an unusual design and higher than now allowed, the Health and Safety Executive report added.
Also, the red signals on the gantry were not directly in drivers' immediate line of vision but slightly off to the left, it said.
In addition, a misaligned track joint could have caused sufficient vibration to interfere with Mr Hodder's in-cab automatic warning system, said the report.
It was "conceivable" that the vibration could have resulted in Mr Hodder receiving a false "green signal ahead" indication as he approached signal 109 which was, in fact, red, the HSE said.
There were "a number of interrelated factors relevant to the potential for driver error on approaching signal 109," said the report from the Health and Safety Executive.
Father-of-two Mr Hodder, from Tilehurst, Reading, Berkshire, who died in the crash, had obtained his train driver competency certificate just 13 days before the morning rush-hour accident on October 5 last year.
He had completed just nine shifts as a driver before his Thames Trains diesel train passed through a red signal just outside Paddington station in west London and collided "virtually" head-on with a London-bound Great Western express train from Cheltenham Spa, Gloucestershire.
Thirty-one people died in the accident in which fire ripped through part of the wreckage.
But today's report made clear that only three people died as a direct result of fire.
The report said: The initial cause of the accident continued to appear to be the Thames train passing a red signal. The reasons why the train passed the red light were likely to be complex and any action or omission on the part of the driver was only one such factor in a failure involving many contributory factors. No evidence had yet been found that the signalling equipment "performed otherwise than as expected". The approach view of many of the signals in the crash area of Ladbroke Grove was complex. The brakes on the leading vehicle of the Thames train could not be tested because of extensive damage, but the brakes on the rest of the train were working normally. Detailed analysis of samples of fuel from both trains had shown that the composition and flashpoints were consistent with what was expected. They were gas oils (diesel) with no added petrol. There would have been no collision had the Thames train been fitted with the automatic train protection system.
The report said that the signals on the gantry holding signal 109 had been designed, in 1994, in the shape of a reverse L with the red signals off to the left.
This was outside of standards at that time, although procedures exist where non-standard designs can be authorised, said the HSE.
The report said: "There would seem to be good reason for believing that driver Hodder did not pass signal 109 simply because he was unaware of its presence.
"The balance of evidence on this would, therefore, seem to suggest that the basis for the incident was one of misinterpretation of the aspect displayed rather than failure to become aware of the presence of gantry 8 (which carries signal 109) and, by implication, signal 109."
On the fire aspect of the crash, the report said two aluminium fuel tanks from the leading Great Western train power car became detached.
These tanks had capacities of 2,710 litres and 1,966 litres respectively and were connected by a flexible hose.
The smaller tank was penetrated by a section of the aluminium flooring from the virtually-destroyed leading coach of the Thames train and it was "likely that this impact resulted in the fuel being ejected under pressure", said the report.
The HSE made it clear that its investigation was still progressing and that there was "still significant work to be completed".
Further studies include the signal design, the effect of the track joint misalignment and how the diesel fuel from both trains behaved during the collision.
This interim report will go to Lord Cullen who begins his public inquiry into the accident on May 10.
Gerald Corbett, chief executive of railway infrastructure company Railtrack said: "This report reaffirms the contents of the previous reports - that there is nothing to suggest the signals were at fault. The height of the signals on gantry 8 accords with current requirements."
He went on: "The signal (signal 109) complies with rail industry standards for signalling sighting. Alterations in 1994 to lower the signal and move the red aspect to the left were carried out specifically to help improve drivers' visibility."
On the report's suggestion that a misaligned track joint could have caused sufficient vibration to interfere with Mr Hodder's in-cab automatic warning system, Mr Corbett said: "The HSE do not say there was a malfunction. They go no further than saying it is conceivable and that this theoretical possibility is now being assessed, however remote this might be."
He added: "Railtrack has only been informed of these matters recently and is determined to get to the bottom of this and expects to participate in any scientific assessment. It has not been provided with any scientific reports to support the theory.
"The equipment has been in use for decades and had the Railway Inspectorate or British Transport Police considered that this was a major railway problem, no doubt they would have informed us well before now."