Agreed, REW let's you decide what to do with you're audio signal. With Audyssey you flick a switch and hope for the best.
Let me give you 2 examples where software like Audyssey has problems:
1. A general rule about tuning speakers is, "do not try to make the speaker do something it can't do". If a speaker has a dip at a crossover point, the software will detect this dip and try to compensate by boosting the eq. But the problem is that the speakers simply do not work well at that particular point, so boosting will just make things much worse. The software do not take notice of the capability of you're system, it just looks at numbers and robots don't have musicality.
2. The time domain, Audyssey states: "In the time domain, the first arrival of sound is seen as a large spike and it corresponds to the sound signal coming directly from the loudspeaker. A few milliseconds later, copies of the direct sound signal arrive after they have been reflected and modified by the room (walls, ceiling, floor, furniture, etc.). Because these reflected copies arrive so close to the direct sound, the brain blends everything together and this results in audible artifacts such as smearing of transients, ringing, and comb filtering. The result is muddy sound. MultEQ filters are specifically designed to address these time domain problems and concentrate most of the signal energy in the direct sound. In the graph on the right hand side, the reflections have been dramatically reduced and the first spike is narrower and stronger thus allowing pure sound from the loudspeaker to arrive at each seat."
To align the speakers on the same time line is no big science (now a days), but to remove room reflections without room treatment - that will properly give a Nobel price
What they can do, is introducing a reversed phase signal or dynamic EQ to the audio signal coming after the "direct sound". And this way achieve a "nice" reading on the graph by noise canceling or dimming the freq. that would be present in the reflected signal. But the price is screwing around with the audio signal big time and losing some of the program in the org. audio signal.