I'm going to write a test to once and for all let me know if any dupes are making it past the stitching. I've almost completely rewritten my stitching code in the last day or so to be much more robust... but hey, it's possible it's still acting wonky.

An interesting note: I also implemented the Melax algorithm, and while it does produce some weirdness of other sorts, it doesn't have holes in the output mesh. It's possible that the algorithm I'm using just has problems.