I'd almost have to see a disc at this point, to analyze the contents in IFO Edit, as well as some other methods of observation. Another choice it to try a DVD+R booktyped to DVD-ROM, and see if that makes any difference. Sometimes it does, sometimes not.
You've done due process, and have a good method.
What was the max bitrate in Reference? I'm assuming you did a VBR encode. The DVD template may have been too high, although I doubt it. But I still want to verify all possible settings in Reference. Are you sure the GOP isn't too long, too?
The multi region codes may also be part of the error. Pick no regions.
What specific error do the players give?
I've noticed about a 2-3x coaster % from some projects I've done recently, where MCC DVD-r 16x were used, so I've taken to scanning 100% of media, even the intermediaries (deliverables to clients, to avoid FTP up/down time). Instead of 0-2% of discs being bad, I'm seeing as many as 5% being bad (or five discs per 100-spindle). Did you test each disc? Note that 5% is still within norms, and tolerable for 1st Class media as per the
"best blank DVDs" review. It's just been a few years (4x days?) since MCC hit the 5% ceiling so much.