You mean, in any given place, the noise may appear on both channels at once? I thought some "edge problems" caused that, so a "noise-generating" point on one channel should be about 1/4 cycle away from the the nearest one on the second channel.
True, my method would fail dead if the noise appeared on both channels at the same moment. (I just assumed that as long as one channel remains in constant state, I'm able to intercept and correctly represent any signal appearing on the other channel (either treating it as a valid rotation or discarding as noise). From your explaination of reasons behind the "noise", it looked like it can appear on one channel at a time at most...?