Arabic Character Issue in 10gDB

Am facing Issue regarding the Data in ARABIC Language.

- Fresh installation of Oracle 10g R2(10.2.0.1.0) Software was done choosing Languages: ARABIC and ENGLISH on 'Windows Server 2003 R2 Enterprise x64 Edition Service Pack2'.
- Parameter 'NLS_LANG' was set in BOTH Windows Registry and Environment:
NLS_LANG=AMERICAN_AMERICA.AR8MSWIN1256
- Then DB was installed using DBCA by choosing Database Characterset: 'AR8MSWIN1256' (supports both ARABIC and ENGLISH), National Character Set: 'AL16UTF16' (as recommended by Oracle).
- Setting of parameter 'NLS_LANG' was confirmed to AMERICAN_AMERICA.AR8MSWIN1256 before Export and Import Processes on both Servers.
- Then Export of data was taken from 8i (8.1.7) DB, Characterset of 8i DB: 'AR8MSWIN1256' and National Character Set: 'AR8MSWIN1256' and data was Imported in 10g DB.

Now when we try to retrieve the Data from 10g DB using Backend or Client Tools(Oracle SQL*PLUS, TOAD, JDeveloper etc), we are getting the Data in junk characters like '?????????', but same Data can be seen in ARABIC Language from 8i DB.

SR has already been logged onto Metalink, they are trying with only Re-installation of 10g Database again and again with different NLS_LANG setting. We are running out of time and Oracle is not accepting Severity of SR to be raised to 1, as this is Develoopment Server.
What immediate action should be taken to resolve this issue.

Cheers ! ! !
-6N