I don't have a stake in XQuery but as a practioner working not just
with markup yet still with enough XSLT experience I've found
XML-specific programming languages generally limiting in that problem
domains I'm using them for would often benefit from the kind of
infrastructure and mindshare that general-purpose programming
languages have, such as APIs for database and network access, unit
testing, etc.
From my utilitarian perspective, XSLT's (and supposedly XQuery's) wins
over general-purpose language where literal XML content with
small-scale variable/expression expansions needs to be produced from
input. OTOH, the more an XSLT program makes use of complex and/or
dynamic expressions to construct output markup, the more other
language options become attractive.