Nor would I.
I believe the encoder's receiver is calibrated to ignore all wavelengths but one, which is emitted by the LED. This would eliminate the problem of false readings, but the main problem, that of finding a suitable encoder wheel, and mounting it to a bicycle so as to have less than .010 inch of lateral runout, remain. Thus the design remains a dead end, albeit one that could be solved with a massive infusion of cash. A couple hundred thou ought to do it.