My only suggestion would be to not use PluralEys. Sync manually.
I can confirm this with XF, and also with P2; it probably affects any source footage that contains multi-mono audio channels. With P2, it's even worse--four channels of the A1 channel!
So, what does PE give you? I've only briefly played with it, and can't recall what it does from a source clip standpoint. Why is what it generates unusable, such that you want to replace it?
What plural eyes gives us is the ability to load dozens of clips from multiple cameras of an event and have it sync them all ready for editing multicam. Since some of the cameras don't have timecode (AVCHD) and some of them are DSLR, it's a major PITA to do it manually. Yes we could do it manually, but having done it that way prior to PluralEyes it can take several hours, whereas Plural Eyes does it in about 10 mins or less (usually less).
When I was on the Mac, PluralEyes seemed to sync everything perfectly. Since moving to the PC I've found that the AVCHD and DSLR files don't sync as well unless I Render & Replace the audio for ALL footage. That works, but leaves me with untold numbers of additional replacement audio files plus duplicate clips in the project. It can easily go from just a few hundred clips to several thousand clips, which makes project loading much slower, a pain to keep moving all the new files between PCs and of course lots more files to keep track of and organize (and reconnect) in the project folder. So, reconnecting (replace from bin) to the original footage works well, except for thi one bug. What a major PITA !!!!
What I've ended up doing it adding a second copy of the footage's audio and replacing just that channel for now. Because the XFs are the long running cameras there aren't as many clips, but I sure hope that Adobe fix this one!
No, I know what Plural Eyes does, but what does it give you as far as source footage? Does it extract/render out the audio? Or are you doing that part manually? Other than making project management a bit more difficult, is the extracted/rendered audio a problem?