It's a math problem. You make some assumptions, and then you solve the problem with respect to those assumptions.
Of course all models of the world are flawed. But some are useful. If you're wondering which restaurant you should stop at on your walk from a to b but aren't familiar with the area, you might assume you should pick the next place better than all previous after walking a third of the way home. That would be a reasonable application of this model.
Regarding your last point: I wouldn't base serious life decisions exclusively on an "optimal game theory" move. That would be nuts! This obviously doesn't perfectly model reality, so why act like it does?
There are other assumptions that this model of dating makes that also aren't well founded. For example, the probability distribution of the candidates probably isn't going to be uniform: a person who has dated one person versus someone who has dated fifty is going to have a lot of biases to the selection that the person who dated only one person didn't have at that time.
What I'm getting at here is that this isn't math giving an answer. There are many different places where a person has to make leaps of faith in order to accept this model. And a person also probably needs to make a definitively incorrect guess to start things off.
So if someone were to be wondering if they are being an idiot for not dumping the person they are with, who they like, for the sake of appeasing this math model? They could be prudent: and it wouldn't be going with what is supposedly going to give them a 37% shot at finding their best match. It would be not breaking up with the person they are dating.