Thanks for your interest in JabRef and for adding the test case. We would love to have a patch for the CitationKeyGenerator so the test wont fail anymore.

Sure, I just add a patch to UnicodeToReadableCharMap and all the tests pass in my local machine. However, since we have to manually add all the special characters in UnicodeToReadableCharMap one by one, if a user enter a name that contains characters not in map in the future, this bug will show up again. Maybe we can come up with a clever way to include all the accents' unicode in the future

Some time ago, @koppor and me were thinking of a replacement for the current Unicode mapping table, but we did not see a systematic in the Unicode table that would avoid mapping each single edge case. But maybe you see a solution? Maybe by checking a range of the index in the Unicode table?

I would appreciate it if you could try the icu4j transliteration methods:
Article talks about php, but icu is available for java and we already use that library (icu4j)

https://unicode-org.github.io/icu/userguide/transforms/general/
https://unicode-org.github.io/icu-docs/apidoc/dev/icu4j/com/ibm/icu/text/Transliterator.html
https://bartvanraaij.dev/2020-10-17-converting-utf8-strings-to-ascii-using-icu-transliterator/

I think for the moment we can merge this and create a follow up issue investigating some alternatives