The fundamental problem here is that Drag and Drop uses a GEDCOM export and import mechanism as its data transfer mechanism. GEDCOM is not very satisfactory for transferring such things as groups, sources, source templates, and sentence templates. RM does try to transfer sources,source templates, and sentence templates when transferring to a new and empty database, but it doesn't even try to do anything about groups. Even for sources, source templates, and sentence templates, RM needs a different and better way to create a new database that's a subset of an existing database.
To drag and drop a subset of an existing database into a new and empty RM database, the best strategy I can think of is that would really work would be to make a complete copy of the existing database and then to delete people from the new database that were not marked for inclusion. RM has stated very strongly that they will never support a bulk delete capability, and I suppose my proposed strategy smacks of a bulk delete capability. To be safe, my proposed "make a new database and delete non-included people from the new database" process would be a single, atomic operation from the RM user interface.
To drag and drop into an existing and non-empty RM database is much trickier with respect to sources, source templates, and sentence templates. You have two existing databases and you don't want to change what's already there in either database. A drag and drop into an existing and non-empty RM database could conceptually be done behind the scenes with a GEDCOM export/import or by copying certain rows from tables in one database to tables in the other database. But in either case, the fundamental problem them becomes the proper merging or non-merging of sources, source templates, and sentence templates on a case by case basis. In the case of sentence templates, such merging would be impossible unless the process created new fact types - e.g., a fact type of Birth2 if the sentence templates for the Birth fact were not the same between the two databases. I'm not proposing that RM do so, but the problem is tricky. In the case of sources and source templates, even an improve Drag and Drop could continue not to merge Sources and Source Templates that are not identical. But identical ones really do need to be merged correctly and automatically.