Are you learning with the remote connected to the PC?

If so, then don't.

These are very strange learning errors. They don't fit any pattern I've seen resulting from failure in getting the correct IR signal from one remote to the other. They do fit the patterns I've seen from Windows latency problems when learning with the remote connected to the PC.

I don't really understand what happens when you learn while connected. Obviously it is still the remote's IR sensor in use. But from failures reported by others it seems that a key part of the initial processing of the incomming data is done by the PC when connected but done by firmware in the remote when learning is done while disconnected. If network or display or other Windows drivers happen to grab too big a chunk of CPU time in the middle of that process data is dropped and you get weirdly garbled learned signals.