If this is a bug with Preview, then that's really, really bad since Preview is a bedrock of macOS and has been for years. However... it sounds like the issue i

If this is a bug with Preview, then that's really, really bad since Preview is a bedrock of macOS and has been for years. However... it sounds like the issue is that FineReader is storing the OCR'd text in the metadata in a way that's not part of the official PDF spec. So, it sounds like Preview is able to open the file by ignoring that metadata and then, upon save, is storing the metadata back, as normal, which then corrupts the OCR data. This reminds me a lot of when people would store metadata like this in MP3 files to include things like album art and booklets. Normal mp3 players would ignore it as just metadata or bogus data but opening it in an audio editor would do this same thing. I'm not sure who the "blame" lies with in this case because Abby FineReader probably is writing this stuff in a non-supported way but Preview really should just ignore it rather than trying to correct it. It's very likely that the OCR text, post-save, is actually bits from the document itself rather than from the metadata.