It would, but handling level needs some care - what level should '1' represent? Would probably need to choose a desired target level, e.g. 75dB/A, so that the result of A*B afterwards would be a flat line at the desired level.
A single full scale sample serves as basis, this waveform may be generated in other software, and imported to REW. Scaling is problematic, but REW has no trouble returning the amplitude portion, however REW doesn't return the impulse response:
This isn't always the case. When two similarly produced impulse responses have A/B performed, a resultant waveform is generated.
I don't know how John is doing the math.
A * B in frequency domain is convolution in time domain, and once again scaling can be problematic in REW implementation, but works.