There are a few characters that are dissallowed in XML documents, even when you encapsulate data in CDATA-blocks.
If you generated the document you will need to entity encode it or strip it out. If you have an errorneous document, you should strip away these characters before trying to parse it.
See dolmens answer in this thread: Invalid Characters in XML
Where he links to this article: http://www.w3.org/TR/xml/#charsets
Basically, all characters below 0x20 is disallowed, except 0x9 (TAB), 0xA (CR?), 0xD (LF?)