I can't quite see how this could be a database problem. First of all, I think my database is setup with the correct encoding, and when I modify the entries by hand as described above, it works.
Also, to clarify: The field tterm is quite correctly inserted in the database. The problem is the field tfirst, which - if I understand correctly - should contain the first word of a multi-word term. I don't know how the first word is extracted - is this done in the database or the php script? In any case this shuld do exactly the same as the plugin code does for determining words in an article. I.e. both should use the same set of word delimiters, and that doesn't seem to be the case (the plugin accepts - as word delimitier, the editor code not).