Hi Ben. Sorry I didn't notice your message--I don't read this forum regularly. I haven't touched 3dretroicor in years but I wouldn't mind taking a look, time permitting. I will email you (I try not to post my email for spammers to pick up). In the meantime, a few thoughts to help you diagnose the issue yourself:

3dretroicor does the plain old RETROICOR, almost exactly like the Glover paper but with a few minor simplifications. If I recall correctly, the main difference was in the physio-to-phase conversion, which you can check by outputting the phase and plotting it along with your original physio data. I've never seen a problem with this stage, but it's possible something about your physio data could be tripping it up. You can also use 3dcalc to subtract your original dataset from the processed dataset to see exactly what changes were made. Maybe there's a clue there.