I just spotted this little tidbit in Programming the Perl DBI on page 299:

The NVARCHAR2 and NCHAR variants hold string values of a defined national character set (Oracle 8 only). For those types the maximum number of characters stored may be lower when using multibyte character sets.

...

Oracle automatically converts character data between the character set of the database defined when the database was created and the character set of the client, defined by the NLS_LANG parameter for the CHAR and VARCHAR2 types or the NLS_NCHAR parameter for the NCHAR and NVARCHAR2 types.

CONVERT(string, dest_char_set, source_char_set) can be used to convert strings between character sets... See the "National Language Support" section of the Oracle Reference Manual for more details on character set issues.

I have a hunch that the difference between the two environments has something to do with the NLS_LANG and/or NLS_NCHAR parameters. (Disclaimer: I am not a dba, so I could be barking up the wrong tree.)