I watched the webcast from the presentation of the report. HAVKOM pointed out that four FDR channels (pitch, roll, ground speed and mag heading) were incompatible with other data after the event started. All these channels are generated by the IRU.
So they made a graph showing both the FDR pitch data and their "best effort estimate" of the actual pitch.
The graph shows that the FDR pitch suddenly indicated up. After 2-3 secs the estimate indicates a pitch down which increases, with -1g negative pitch acceleration.
To me, this would fit a scenario where the PF sees a pitch up on the PFD and reacts by pushing down and trimming nose down (as also recorded on other FDR channels). Overspeeding occurs after approx 15 seconds.
Note that the investigators did not confirm such a scenario, but did not deny it either.