What are the actual precision limits of the FPU implementation?
If the issue is that it isn't the full 80 bit but is at least correct for IEEE 754 64-bit, then it's certainly no worse than the emulation used in OS4/MorphOS.
If the issue is that its less than 64 bits, then sure, that will cause problems for some software. How much of a problem will depend on the task at hand. It also depends on which bits are sacrificed. Most likely the low order bits of the mantissa, of course.
If the issue is that it's less precise than 32-bits, then the statement that its just a "quake enabler" may be fair criticism. Although it will also enable many other applications too.
My understanding is that its "not quite 64 bit IEEE" due to some current limitations imposed by gate counts.