I know that this is documented in the changelog and that there are explicit notes and alternative data on the download pages - eventually the map maker is to blame. But I think it's still problematic to have such 'de facto' boundaries in a default dataset that is used by many non-professionals.