For Office Import:
Proposed change vs 4.1.2 is attached in comments 11.07.2012.
Code is adapted from HTML Cleaner library
I cannot test the code, i am not Javist and cannot have all the infrastructure needed to compile run and deploy. But i believe that code is correct or only need minor fixes to be compilable
For Office Export similar change is attached as
Russian encodings fails when:
a) use office import
b) use office preview for attachments
but when use PASTE function - russian encodings shows normal.
Interest, that Romanian encodings shows normal in any case ...
All XWiki+database utf-8 configuration already made.
My config: Win2003, OracleXE, Glassfish, XWiki 2.5.1