Having read the report now and digested the added facts I suspect the most likely story is that the pilot either decided on a manual descent or mis-programmed when setting the altitude preselect and then faffed about fixing the issue and ended up disconnecting the autopilot. From that point it appears a standard loss of control.
My logic is
1 - 'Sam' who appeared to know the pilot personally (earlier in this thread) said he normally flew on autopilot, the BEA makes this statement
2 - Radar tracks do sometimes have the kind of small discontinuities seen in the bulk of the track (look at the G-EYES radar track)
3 - At the bottom of the track the aircraft clearly comes off the autopilot, with a nearly 30 degree turn to the left, then back on heading, then what looks like a 45 degree turn to the right, followed by the loss of control.
This last event happens just when the aircraft should be levelling at 3000 ft, probably just after transitioning to IMC. If one was planning on having the autopilot fly you to your destination, any unexpected manual intervention would be high stress.
As an aside, this accident report does clear up one of the issues that was hotly debated in 2008 - Do the French consider themselves to be part of the Country EASA/JAA? And of course they do not and have affirmed their understanding of 61.3.