I have a dat file that contains setnames with umlauts. When i set my system to japanese environment the umlauts do not get displayed (they appear as a blank) and, what is more annoying, the sets are ignored. They get neither removed nor scanned. (Although clrmame says, it removes the 'unneeded' sets, they are still in rom folder).
Well...you discovered the problems which arise with non-standard characters in datfiles. By default OEM/ANSI conversion is done when reading zipfiles.
Problem is, such conversions are depending on code-page settings. What you can try is to turn off (or turn on...depending on what you try) OEM/ANSI conversion in Settings->ZipSettings.
The better method would be: Kick the datfile author 8)