If I understand correctly you just cluster your files, than hit save, right? Clustering by itself does nothing to your files, it just uses existing tags in your files to guess which files belong together. When saving it just uses your already existing tags to save and rename, and if there are no existing track number tags the saved filenames also can’t have track numbers.
What you actually should do after clustering is selecting the clusters and then use the “lookup” button. This will search for releases on MusicBrainz matching your clusters, load the data for those releases into the right pane and move your files from the left pane in that right pane matched to the loaded releases. You can then check if everything was matched correctly and save.
General rule: The left pane where you also see the clusters just holds your files with existing tags. You can manually edit tags there, but as long as the files stay in this pane Picard will not use anything from MusicBrainz.org.
The right pane shows data (release and track metadata) loaded from MusicBrainz. Only if you files are matched to those track will they use he new tags from MusicBrainz when being saved.