That nose-down command between 2:10:16 and 2:10:50 reduced the vertical speed from 7000 fpm to 700 fpm and AoA to (less than - in my simulation) 4 degrees (reducing v/s implies less than 1g, hence reduced AoA) - no stall warning at that point. Then a nose-up command must have followed calling for about 1,1 g and an AoA that exceeded the stall warning threshold, and ultimately the stall AoA. The 1,1 g increased the vertical speed from 700 fpm (to about 2500 fpm in my simulation). Look at BEA's "3D view" for the partial level-off followed by a steepened final climb just before the apogee. The airplane stalled before the apogee.
The BEA text, and the "3D view" graphic, seems to me to indicate that stall warning was before the second nose-up input and ascent. It is almost as if PF reacts with back stick every time stall warning goes off (even before it starts doing strange things below 60kts). We'll find out, hopefully, if that is really the case when we get more data in the next report.