Here's a real stoopid hypothesis based on my distrust of computers.
Many a time we have seen competent pilots put their faith in what the pretty computer graphic is displaying and initially ignore warning lights and gongs generated from completely separate sensing systems.
Can the main tank qty display get out of calibration and read 0kg left when there is in fact actually 75kg available? Can the main tank qty reading go backwards without a reset after registering a 0kg reading?
Pilot switches off XFER pumps when 0kg/main is shown on the pretty computer graphic unaware that the logic that drives the fuel qty display has gotten all confused and produced an incorrect indication that eventually conflicts with the final safety net of caption and aural alerts.
Pilot knowledge of the separate circuits for fuel captions and aural alerts is probably up to date however familiarity with endurance of the machine compels the pilot to believe that their is fuel available, so pilot makes the assumption that the captions are probably wrong and ignores the FM procedures to land immediately, after all that pretty computer graphic is so convincing, not to mention home base is only a few minutes away.
Unfortunately this hypothesis doesn't explain why after the loss of the first engine the pilot didn't put two and two together and immediately lose all faith in that pretty computer graphic.
This is the bit that is really hard to understand, but we have all heard and read about these kinds of incidents of continuing in the face of multiple warning indications many times before.