In monuments_all table, we do not keep the origin wiki project of the monument.
The ErfgoedBot categorization bot tries to match an image to Commons categories, using several methods.
One follows the monument_article to look for {{CommonsCat}}-like template on it. But this assumes that the monument_article lives on a Wikipedia project. As we have lists from other projects (like WikiVoyage), that the harvesting job supports, then we cannot categorise these files.
Update: There is now a project field in monuments_all. What is needed is for the categorisation jobs to make use of this new data.