There are two ways I can see your argument going.

• Physical laws are descriptions of a universe that has its own (potentially inaccessible to us) rules.
• Physical laws are descriptions of a universe that has no rules at all.

In the first case either we can write down the actual rules of the universe inside the universe, or we cannot. Whether the rules can be expressed within the universe as correct natural laws depends on its nature. Given that the rules of a system as simple as arithmetic can be fully described within the system and can be used to prove facts about its nature I think it's at least plausible that we could express the actual rules of the universe within the universe, which would demonstrate a counterexample to your argument. If we cannot express the actual rules of the universe then your argument is correct and only approximate descriptions of the rules can be created and we should use descriptive natural laws with the most benefit for a particular purpose. Amusingly, if the actual rules of the universe are inaccessible then there is no way to rule out immaterial souls or other supernatural phenomena. The inaccessible rules may or may not allow the supernatural, and we will never know.

In the second case our universe happens to exist as if it had rules entirely by chance. Perhaps there are an infinite number of universes where every possible configuration exists and so we (as finite decision theory machines) only experience the universes that appear to have rules. Or perhaps we exist as fleeting instants of time across a sea of an infinite number of universes, our memories and experiences linked only by the similarity between different random patches of universes.

There are a few inferences we can make about how we should be rational depending on which case we think we're in. In the first case we should believe in induction, e.g. the universe probably has mostly-constant rules we can rely on in the future. In the second case one might think induction is invalid and we should have no expectation that the universe will exist or be meaningful in the future, but actually our very existence depends on rule-like formations of the universe existing. We will experience a future instant whenever a random portion of a universe matches what we would expect the rules to predict. A universe that happens to get all the atoms lined up just right to produce an instant of our planet's history will, by definition, line up the atoms very closely to what our current natural laws predict a universe with rules would do. If enough atoms were out of place we would not exist or experience anything. In this sense the natural laws we know would not control the behavior of the universe so much as they controlled (or defined) our continued existence, so behaving as if natural laws existed would still be rational.

There are philosophical theories, modal realism, and mathematical realism, that propose any set of natural laws and initial conditions describe and cause an independent reality, or alternatively stated that any possible universe or world independently exists in reality. The difference (if any) between the class of possible universes and the class of universes defined by natural laws and initial conditions would be the class of universes without rules. The obvious problem is that "possible" is such a nebulous term that it's hard to know what it means. Maybe it means universes that cannot be described with ZFC set theory would qualify as possible universes without rules, or maybe it means that ZFC is insufficient to categorize all possible rules (the latter is more likely, in my opinion).

Intuitively I think we are in a universe that can be fully described using ZFC, since it seems to work so well for modeling what we can experience so far. The question is whether we can derive that full description sometime before the heat death.