1) I tested also jt400-9.1.jar from the jtopen-package. It made no difference
2) The iSeries Release is V7R1M0
3) The iSeries-"DB" is not really a database, only a collection of files with a SQL-Frontend on top of it. so every file has its own encoding (CCSID). The system value for CCSID on this machine is 65535, but the files have normally 37
I looked through some more documentation and websites. From the JDBC FAQ:
Why is the Toolbox JDBC returning EBCDIC characters to my Java program?
Normally, the Toolbox JDBC driver will translate EBCDIC characters to Unicode Strings automatically. If it appears to be returning untranslated EBCDIC characters instead, then it is likely that the field in the IBM i database is tagged with CCSID 65535. The Toolbox JDBC driver recognizes this CCSID as a field that should not be translated. To avoid this behavior, tag fields on the IBM i that you want to be translated, with a valid CCSID. Alternately, you can set the "translate binary" connection property to "true", which instructs the JDBC driver to translate all fields, including those tagged with CCSID 65535. The easiest way to do this is to add
to the end of the URL used when connecting to the database.
I added this property to the advanced parameters of the connection and my problems are solved ! :D