It seems to be somewhere within the xerces xml parser library so i guess, the change of the test ignoring the thousand delimiter character should be ok.