The way I think of it is "Above this RPM use filtered MAF for airmass
(in place of) air mass prediction calculation." As stated by HPT's description "for" means the same thing, but it's ambiguous.
Airmass prediction calculations account for transients with high accuracy. This is due mostly to the comparison being drawn between MAF and MAP across the throttle body. (Details here:
https://forum.hptuners.com/showthrea...ynamic-airflow). Thing is that's also computationally expensive. The thinking is that below 3900 rpm there is a lot of on-off-on pedal. That's where most normal driving occurs. Above 4000rpm it's usually a full power pull (acceleration) or holding (heavy towing), so airflow changes at a more or less linear rate. Again, due to compute resources, the compromise must be made to slim down calculations in order to keep engine control at the high resolution required to ensure safe power delivery.
Those enables could technically be set to run full-time. The problem in doing this is that at higher RPM's the synchronization of all calculations, fundamentally based on crank/cam triggers, will happen at a slower rate. Things that might need corrected or might need flagged could slip between loop updates, being addressed later than optimal.
Setting those values too low, disabling the airmass prediction, reduces accuracy.
So, air/fuel accuracy as much as possible and then transition to filtered MAF in order to maintain control as data and computation load increases.