The Saxon processor is a full XSLT 2.0 implementation. Look for the
unparsed-text function in the docs.
It's basically used in the same way as the document function but the result is a plain string not an XML document.
It means you can bring in text documents, you still can't have a non-XML document as the primary source of the transformation.
I was fairly sure that there was also an example in the book you mentioned but perhaps I'm wrong.
Added: I think the example mentioned is in Chapter 8 in the XSLT 2.0 section. An old fashioned ini file is transformed into a more modern XML representation.