boundary=religious_administration occurs mostly in Poland. Do diocesan boundaries there conform to civil boundaries? If not, that might have been a big reason why those were mapped there. Even so, the Catholic Church has dioceses in very many countries, but the absence of them in OSM from most countries speaks to the community’s reluctance to maintain data about them directly in the database.
May I ask about your particular use case for constructing these diocesan boundaries directly in the OSM database, as opposed to a mashup or query based on OSM? For what it’s worth, it’s possible to join OSM geometries to Wikidata, which has extensive coverage of dioceses. The big caveat is that, on Wikidata, only a couple of items about Roman Catholic dioceses in the U.S. currently indicate which counties or states are served by a given diocese – namely, the two that I’ve tagged. 
Kind of underwhelming, I know. Wikidata has much better coverage of this sort of thing in some other countries. It’s just a matter of adding operating area (P2541) or coextensive with (P3403) statements to each item. If you have an external dataset of machine-readable information, we can automate this process.
For what it’s worth, OpenHistoricalMap has started to directly map the boundary evolution of Catholic and Jewish ecclesiastical boundaries in some countries, including diocesan and parish boundaries, so your contributions would fit in quite well there. Diocesan boundaries in the U.S. be especially interesting on a time-sensitive map, since they weren’t always based on county lines. More importantly, OHM doesn’t generally expect features to be directly observable, so this wouldn’t be “exceptional”. It wouldn’t be unusual at all for you to map a boundary based on consulting some external reference.
(That said, OHM would probably refrain from mapping pizza delivery service areas, but only until someone happens to care about a particular pizza parlor’s history to that extent. It would be a good problem to have. 
)