Consider the difference between "vanilla" natural selection and sexual selection.
On paper sexual selection sounds strange; while straightforward selection says that a trait spreads because it makes an individual better at life meaning they'll have more descendants, sexual selection says that a trait spreads because it makes an individual attractive to potential mates, resulting in more descendants sure but it suggests that "being better at life" isn't required. And indeed sexual selection is often used to explain traits that make an individual worse at life, like the peacock's tail.
What's going on is that there is a first-order and a second-order process of causality, that are related but distinct. Say for instance you have a trait that makes your offspring 10% fitter. Awesome, they'll be 10% fitter. Except that this means the offspring of your mate will also be 10% fitter... which means other individuals have an incentive to be your mate, meaning you can get a better mate who will also give fitter offspring, so actually you might end up with offspring that are 50% fitter. The latter process is sexual selection, and while it's dependent on the first process to happen in the scenario I described, the causality, mechanics and results are different. You can have one without the other; for example if a trait makes you 10% fitter but this is completely undetectable to anybody else, they won't be able to select you as a mate on the basis of it so you won't have sexual selection. On the other hand sexual selection can indeed in theory function with something that doesn't improve fitness on the first order.
You can see the sexy son hypothesis as a third-order distinct process. It's not just that you have an incentive to mate with a person because your offspring will be fitter; there is an additional incentive involved in the fact your offspring will also be more attractive. This isn't a tautology because it isn't a foreordained conclusion; for example if you have a "love red hair" gene but nobody else does then you'll pick a redheaded mate on the basis of their sexiness, but your sons won't be sexier for it. On the other hand if everyone except you has the "love red hair" gene, you have an incentive to get a redheaded mate even though you yourself don't find them sexy (or not for the same reason everyone else does at least).
This process certainly allows for sexiness to have no fitness-based cause but it doesn't presuppose it; indeed one would expect for there to have been a fitness-based cause at the beginning of the chain if not at the end.