"APOLLO HardFPU is 100% identical to the 68060 FPU from instructions."

This can mean a lot of things. For example, it could mean "double precision only". This is probably ok for many applications. However, if this is a 68060 FPU, it also means that the transcend functions are missing. This is per se not a problem, there is a known good, known working solution for it. However, if you want to compute a function such as "sin" with double precision, you need some extra precision on top for the internal approximation. The Motorola FPSP uses extended precision internally where necessary, and ensures double precision output.

I do not know how to get a double-precision-precise sin-function without extended precision math. sin() is slow anyhow, so it would not be necessary to have an extra-fast-math for this.

Then again, to be fair, I do not know which Amiga software exactly required this either... I'm probably too much a mathematician here.