[Koha] Character problems

I've been trying to import some records to Koha 2.2.9 system (both with Koha's internal breeding.pl tool and bulkmarcimport.pl), but I can't seem to get all the UTF-8 characters to work right. While breeding.pl seems to break nearly all the special characters, bulkmarcimport.pl does a much better job. There are however some problematic characters with bulkmarcimport.pl also. For example a with umlauts (ä) works, but followed by n (än) it becomes a mess, very strange... I've been thinking about doing a database-wide search-replace to fix the broken characters, but how should I proceed with that? Is it a very good idea in the first place or is there a better way to fix the characters? Thanks! -Pasi-