If you have already used XTM, an online CAT tool, you should know that there is no chance to obtain (download) what you translated.
Furthermore, its concordance feature is below average IMHO, and you have no control on the glossary feature (you cannot add terms)
It means that if you have to work again on the same project, and you aren't so lucky to have the same memory of Pico della Mirandola, maintaining the terminology coherence becomes a nightmare and a waste of time.
Now, the only downloadable file with both source and target is a .xlf file, but it's quite awkward to index with Archivarius 3000, as spaces and full stops are killed making text hard to read.
Luckily enough, an Okapi Rainbow (Rainbow is a graphical user interface to launch Okapi compliant utilities) tool makes the job and does it for free.
The utility is XLIFF Conversion and it turns any .xlf file in a TMX file
Note: the suffix of converted file is again .xlf (indeed out.xlf) but don't panic: its look under Archivarius 3000 is entirely different, thanks to spaces and full stops that are all retrieved