I built a more rigorous sandbox to check maintenance costs, all the terrain is up to date, attached below.
We have 4 choices for second city:
A - costs 2gpt to start.
B - 3gpt to start, settled a turn later.
I think I might still go with A, even with a guaranteed religion. A builds up a lead until we invest 20h into the first WB, then only starts to catch up when we put another 20h into the second WB. A LH becomes a nice investment then, but hopefully that'll be the case for some later city.
C - 4gpt to start, settled a turn later, saves 3 worker turns, but I'm not sure what to do with these.
D - 5gpt to start, turn later, can't start improvement immediately.
These require turning one of our scouts around pretty soon, which is rather annoying. Our workers don't have to waste turns moving to the other side of the capital, which is nice ... if we can find something for them to do. Another gpt down on the wheat tile too, that more than makes up for the gems really. Even with a guaranteed religion it's debatable IMO, so A is probably better on the balance. They're good locations though, even if we have to whip in a monument.