Well that comment is a head scratcher, but never mind.  I do this pretty much every day in my day job, without the merest shred of XQuery.  As a matter of fact it's pretty interesting that in one of my recent projects I led development of a particularly intensive multi-hierarchical project in pipelining code in Python (the target corpora was represented in XML).  Another group re-implemented it in XQuery, for reasons I fully understand. However, after taking one look at the XQuery the greater update by other participating groups has been through the Python code.

Others could choose differently, of course, but it just shows that there are indeed many ways to skin a cat. Some are actually quite straightforward and others rather less so.

Oh and when degeneration into tabular data is required, you need a good information architect, regardless of what technology you use, because the entire problem space moves from bits to neurons.