odbc silently dropping national characters in CP1250 encoding

I have a LATIN2 database on the server side and a Postgresql ANSI odbc
connection on the client side (windows xp sp2, latest odbc driver,
version 8.03.04.00)

when I try inserting something like this:

insert into sample_table (sample_column) values ('öüóőúíéáűÍÜÓÖŐÚÉÁŰ')
insert into sample_table (sample_column) values ('čćšđž ČĆŽŠĐ')

and then selecting it, I get the following:

öüóoúíéáuÍÜÓÖOÚÉÁU
ccšdž CCŽŠÐ

So, the characters ő,ű,Ő,Ű,č,ć,đ,Č and Ć have been silently replaced
with o,u,O,U,c,c,d,C and C

the same thing happens when database is created with WIN1250 encoding.

when I try doing the same thing from pgadminIII, everything is OK -
characters are inserted and read properly so I suppose the problem is
somewhere in the ODBC connection. Am I missing some setting or how can
I set this up properly?

Dragan Matic