AnFI: two points to ponder from the video.
1. The last second or so before impact is, due to change of pitch/nose attitude, something like a flare (back cyclic) which tends to increase Nr, see crab's point on that.
2. The aircraft was flying in a nose down descent after the turn/course reversal, but it was powered so the Nr would be expected to have been within normal range until the flare at the end that immediately preceded impact. What this means is that once the maneuver began, through its course reversal, and as it descended and regained speed, airspeed was not stable/constant ... and ... at that flare to level from nose down and the last pull to stop descent (too late!) airspeed was not constant during the period during which you are trying to establish 90 knots as a quantity of relevance to the event. I presume that during the descent some collective may have been increased before that last pull, but without in flight data cannot know for certain.
Just out of curiosity, when is the last time you flew a rotor over like the one the Apache attempted?
If you are going to model a flight phase, or a rotor disc phenomenon, trying to do so in the process of a dynamic maneuver seems an odd place to start.