Let me try to make my point another way. Suppose the signal from the source is sent to the AVR over HDMI and then recorded on a harddrive in the AVR. And, at a later point in time, the AVR reads the signal stored on the harddrive back into memory and then sends it out over HDMI to the display.

Can you see in this case that provided the signal arriving at the AVR could be parsed into 0s and 1s then timing imperections, etc., from the this original stream to the AVR would have no bearing on the stream sent from the AVR to the display? It wouldn't matter if the voltage level behind the 0 or 1 was nearly perfect or almost indetecable - in either case, it would be stored on the harddrive as a perfect 0 or 1, and then read back from the harddrive accordingly. (Or, at least, however imperfect the 0 or 1 was stored on the harddrive would be the result of the harddrive circuitry and not the HDMI input circuitry.)

This is the basis of my argument - that just becasue the AVR receives a data stream and then sends it out, that doesn't mean that the AVR is repeating the electrical/voltage/timing anomolies from the input to the output.