What sort of characters are you seeing? Are just the delimiters looking like EBCDIC? It would be strange that only some of the characters are EBCDIC.
It might be that the delimiters are in the range that TN has trouble with, which I think are in the 128-135 decimal range. If the delimiters are in this range I believe there is an encoding bug in the EDI processing that causes the validator to barf. If this appears to be the case (intercept the file before it gets into IS/TN and view the delimiters with a hex viewer). If this is the case you might ping tech support to see if there is a fix.
#B2B-Integration#Integration-Server-and-ESB#webMethods