That would have been ideal. But obviously the problem wiki data was not a part of the source wiki in your last test of the import. You can either test again with this new source data, or try it live.
Personally, the only problem I noticed with the edits above involving trailing zeros, and I gave the fix for that. I have tested with a few articles that contain non-ASCII characters, and I did not see any problems; for the characters that were considered valid, they were left untouched. But I cannot test and review that it works for all 1,112,064 characters in the UTF-8 character set. That is almost beyond the testing anyone can do, even you if you had another test import to try. I do know that it works for the invalid string cited above "\xB1-meth..." and correctly changes it to a valid string as "?-meth...", and that the same regex used by this cleaner is used by other applications without reported issue.