I'm having trouble guessing why those bad learns didn't work (now that Jim corrected me about RF but didn't give a new theory why the signals wouldn't work). I understand IR signals very well. I don't understand Prontos well (especially the variations by model).

phatstadt, when you tested those signals was each the only action on a button or were they used in a sequence? Bad learns like those might fail to work when used in a sequence; But they shouldn't be bad enough to fail as the only action on a key.