PrinceXML is excellent, we used it for many years for typesetting programs for conferences, etc. After a while, though, I picked up enough LaTeX to be competent at generating LaTeX from our applications (just as we would generate HTML or any other markup) and that's what I use for report generation now. While there are more templates to maintain, the quality of the results is absolutely worth the effort.
On a tangent, a very nice "win" with the (pdf)LaTeX approach is that it's fairly easy to collate other PDF documents into a larger one, with a custom table of contents, consistent running headers and footers, index of authors, etc., and intersperse them with other content from our database. (This was done by splitting the child documents into pages with pdftk, then inserting them as "PDF graphics" into the generated LaTeX document -- and as long as the original PDFs were searchable, they remained searchable in the final output.) In one of our paper-intensive processes (a proposal approval workflow), this approach cut down tremendously on manual labour that was being done to prepare proposal dossiers, and the resulting collation was much easier for the reviewers to work with (e.g., everyone could turn to page "K-34" in the dossier, and be certain they were looking at the same page of the same document).