I once read one of the mythical titans of Civ3 (i.e. SirPleb or DaveMcW or someone like that) propose the simple rule of thumb that during REX you want to maximise population growth, so if your next city site can match or surpass your capital's +fpt, then go settler first. Otherwise, effectively doubling your capital's +fpt with a granary will be better.
**Edit - I failed to mention that all of my current tests have taken place in an environment where there is no bonus fpt available at the capital site and the starts all begin with settling in place - no moves allowed.**
Not that I want to put myself up against any titans, mythical or otherwise . . .
In my tests (now 15 of 36 complete) I so far have not found this rule of thumb to be true. It seems to make sense, but I have not yet been able to make a settler start expand faster than a granary approach. Even the worker first approach results in a better start - (generally) a shade behind the Granary approach. And as I added food sources, the settler start actually drops further behind in both cities and total population. I am starting to wonder if you will only go with a settler first if you are trying to escape a shield poor desert start. Of course you may use a different opening move depending on your VC, nor am I saying there is only one way to play the opening moves.
Talking PTW, the worker first approach actually has produced better results (again, just a shade better) with an Industrious Civ if there is a bonus fpt at a second city site.
One other thing I noted as I have added bonus fpt is that corruption becomes a deciding factor (I know that templar_x already noted this, I am just validating).
The particular problem is that the city with the bonus fpt attracts the workers but some of that work is lost to corruption. When deciding if the capital or the 2nd city uses shared improved tiles, I was defaulting to give the 2nd city the good tiles to try to create a settler factory. I think I need to test more to find what would happen if I made the capital the priority and see if it has better results. For a straight REX point of view, I don't see it would, but that is what tests are for, right? The logical prediction is that REX slows (fewer cities/citizens) but it will create more military and higher gpt for other things (like research).
And I have a retraction based on templar_x's warning - maximizing research during REX is definately tricky. Just based off available gpt and treasury, other than in the 'control sample', the settler first approach has yielded the highest available gpt and highest ending bank of gp. It is way too early to call this one. Templar's observation that REX and research are inversely proportional (to an extent) may be unavoidably true. But the further behind the settler REX falls, the higher it rises in gpt/treasury. The difference also appears more pronounced in the non-industrial civ. Again, way too early to call. Entertainment costs are the huge factor here.
Another thing I tried was irrigating all bonus fpt under the belief that food is the one thing that is not corrupt and population=power. I found that it a) did not work well and b) required a LOT of MM to make it worthwhile. Needs more testing, but I wince at that thought. Main problem here is that a lot of irrigated bonus fpt= lots of citizens, but no shields results in a lot of large cities that demand 40% of your income in entertainment. Pop rushing might be the key to solve that issue. I'm not great at pop rushing, so that will be a new adventure. I’ll need to re-read Cracker’s Opening Moves for sure.