... its unmatched expressiveness and conciseness when manipulating XML documents.

You obviously haven't tried doing anything other than very simple translations and substitutions. Try doing something that expands one element into a number of element groups; or collating a large number of groups in order to produce summary statistics; and see any elegance (which I never perceived), and any pretence at conciseness disappear.

But mostly you are missing that an XML document is never the final form required by an application. Try loading a wordprocessor document from it's native (binary) format, and the same document from an XML representation and see how much extra time/cpu is used by the later. Same thing for spreadsheet data, and rdbms data, etc.

Try performing SQL type outer, inner or cross joins on tabular data held in XML format using XSLT. Then tell me that it is elegant and concise.