Ok, I averaged per second, also to get rid of the stretch of diagram-time when many values per second. I know it's still not perfect. I also removed datasets with no_position=N
You can see that the ALT drop does not flatten out at the end. Nor is it ballistic after 07:40:17 UTC.
Somebody mentioned stallspeed warning at 4000ft. Don't forget, to deduct IAS from Groundspeed, and Stall from IAS is impossible to guess. And in case the plane nose dived, you have variable GS over descend angle. Forget the Speed. Only the GS increase at the end is somehow a mystery.