What XSLT processor are you using?
The serializer should never output numeric character references (like & #65535; [I think ampersands are getting lost in this forum]) within a CDATA section, because XML doesn't recognize them there. This looks like a bug in your processor.
Secondly, Unicode 65533 is a substitute character for use when a character is found that can't be output in the selected encoding. If the selected encoding is UTF-8, I can't see any reason why it would be used.
The first thing to check is that your input XML is correctly encoded. What is the actual encoding of the c-with-cedilla (use a hex editor to find out), and what is the encoding specified in the XML declaration of the input file? Do they match?
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference