That is because I am not referring to the HTML style-sheets,
that was made quite clear in my original email. It turns out
that the XSLT spec states:
"CDATA sections should not be used except for text nodes that
the cdata-section-elements attribute explicitly specifies should
be output using CDATA sections."
So is your problem that the text did in fact show up but it was not
enclosed in the CDATA section? If so, you should never have expected
that it would be. CDATA sections are syntax sugar, nothing more.
Processes should not and in general do not depend on whether text is
enclosed in a CDATA section, escaped with character references, or
escaped with entity references. It's all the same thing. If so, then
there really wasn't a need to "go through re-inserting the CDATA
sections."