@Will, looks like we are in the same boat, since I've done all those test-individual-parts-before-you-put-em-together, and I seem to have a similar issue with RF decoding, not so different from IR decoding.
See my current open post here
Everything else being similar, my main attention at the moment is around clock's accuracy, since such decoding logic depends on accurate timing. While there are some tolerances built into the algorithm, with sufficient clock drift/skew, it can still be thrown off, and that is what I suspect is happening. I know my RF decoder algorithm inside out, and given all the other testing I've done around it, I see very few other possibilities... may be, barring the possibility of INT0 handling, not being exactly similar as one expects in 328p based Arduino, although possibility of that should be rather lean, since the logic for hooking an ISR to external interrupt should be a pretty lean-n-thin one.
I've got one suggestion to check/correct the clock accuracy issue thru calibration, and found CodingBadly's TinyTuner, but missing bits-n-pieces of how to rig-it-up. Details in my same thread, and also this one
(which is more about the harness for calibration).
I shall be watching this thread as well.