In some cases MLang (on which MSXML6 depends) can added extra ? to decoded UTF-7 data, which can cause UTF-7 encoded XML to fail to parse.
UTF-7 isn't a great encoding anyway, so this is just another reason to Please Avoid UTF-7.
In particular there doesn't seem to me to be much reason to use encodings other than UTF-8 or UTF-16 with XML data. XML is new enough that Unicode support exists for whatever the XML is being used for.