Experiment time! This is done in 474 as Rome with no mods (other than modding the savegame for controlled tests).
Roman Religion - Control Test
"Roman" religion provinces: 9
"Roman" culture and religion provinces: 3
"Roman" religion population: 279
"Roman" culture and religion population: 117
"Roman" religion average civ value: 66.7
"Roman" relgion highest civ value: 90
Temples: 1
Religious Power: 5.4
Test 1: Change population numbers
"Roman" religion provinces: 9
"Roman" culture and religion provinces: 3
"Roman" religion population: 250
"Roman" culture and religion population: 117
"Roman" religion average civ value: 66.7
"Roman" relgion highest civ value: 90
Temples: 1
Religious Power: 5.4
Conclusion: population numbers have no effect.
Test 2: Change number of provinces
"Roman" religion provinces: 0
"Roman" culture and religion provinces: 3
"Roman" religion average civ value: 66.3 (minor change here, will test it separately anyway)
"Roman" relgion highest civ value: 90
Temples: 1
Since the population test showed no change, the rest of the numbers have been ignored.
Religious Power: 5.4
Conclusion: very unexpected result, I thought it would change at least a bit...
Test 3: Change number of culture provinces
"Roman" religion provinces: 9
"Roman" culture and religion provinces: 2
"Roman" religion average civ value: 66.7
"Roman" relgion highest civ value: 90
Temples: 1
Religious Power: 5.4
Conclusion: kind of expected this one. No change, again.
Test 4: Change highest civ value
"Roman" religion provinces: 9
"Roman" culture and religion provinces: 2
"Roman" religion average civ value: 64.45 (again, minor change; will test separately)
"Roman" relgion highest civ value: 70
Temples: 1
Religious Power: 5.4
Conclusion: damn, another unexpected result!
Test 5: Change average civ value
"Roman" religion provinces: 9
"Roman" culture and religion provinces: 2
"Roman" religion average civ value: 10
"Roman" relgion highest civ value: 90
Temples: 1
Religious Power: 5.4
Conclusion: this is getting old.
Test 6: Change number of temples
"Roman" religion provinces: 9
"Roman" culture and religion provinces: 2
"Roman" religion average civ value: 10
"Roman" relgion highest civ value: 90
Temples: 9
Religious Power: 5.4
Conclusion: I give up.
Overall impression: religious power is borked.
So, out of curiosity I decided to compare the religious data from 474 to that from 722.
474 (1 Roman temple):
722 (4 Roman temples):
I'm so lost. Hello, broken feature!
Edit: Ahh, I did another quick test, this time actually continuing the game from the moment the changes were made. It looks like religious power grows steadily after you make changes; so rather than it being a hard number derived from provinces, it actually just grows over time. It would be fascinating to take a closer look at this over the course of a long-term game.