What strikes me most about this sort of discussion is the type of graph you often see. Rolling resistance vs Tire pressure. Initially the resistance goes down as pressure increases. Then there is a break point and it goes up. So it doesn't seem unreasonable to look for that break point and get the least rolling resistance.
But when you look at the shape of the graph it curves gradually into the break point then sharply away from it, as pressure increases. In other words there are diminishing returns as you get close and then a more significant penalty when you go over the break point.
I agree there is some good info on the Silca site. For instance part-4b-rolling-resistance-and-impedance
has the type of graph I'm talking about. There is also a chart that puts the above/below break point thing into example figures. In their example there is 1 watt penalty for being 10 psi under the break point but a 6 or 9 watt penalty for being 10 psi over (depending on the surface).
I don't know what sort of surfaces you are racing on. In the UK a triathlon, in my experience, is often on not great roads. In the worse case I wouldn't get a chance to ride the bike route before hand and it will be on a lot of different roads and therefore surface. So what surface would someone plug into a calculator to get the recommended PSI? You'd have to use the worst surface that is a significant distance.
Given all this my take away is not to sweat it.
Your tire will have a min/max pressure. They make the thing and come up with those numbers for a reason. If you pump your tires to the min figure then you will be the better side of the breakpoint. My reasoning is that only using the resources of something like Team Sky would someone have an accurate breakpoint figure and what will they get? 1,2 maybe 3 watts on me?
The counter point to that is if you do get a chance to ride the course beforehand with time to experiment then I think you can feel the break point through test riding and gradually pumping the tires up. But that's another conversation but I'd rather do that than type in guesstimated variables to a calculator.
TLDR relax, take a safe low guess and it'll be fine.