Does Science Produce Knowledge?

The following seem to be an inconsistent triad:

1. The ideal gas law is a paradigm case of scientific knowledge.
   
2. All items of knowledge are true propositions.
   
3. The correspondence theory of truth.
   

Since the ideal gas law, taken as a literal statement, either assumes the existence of things that don't exist or doesn't apply to anything and is thus empty, the ideal gas law is cannot be a substantive, literally true proposition. Either we have to deny that this is a major achievement of scientific knowledge, or we have to deny that truth is a necessary condition of knowledge, or we have to deny that an idealized law about non-existent entities is false.

For (1), you could substitute almost any bit of science you like, and it still raises the problem. E.g., Newton's Laws are a paradigm advancement of scientific knowledge. Copernican astronomy is an advance in knowledge over Ptolemaic astronomy. Most of science is either heavily idealized, has been shown to be (strictly speaking) false in the light of more recent theories, or will likely be shown to be false in the light of future theories.

One might consider various realist ways out of the problem. So, structural realists say that there is a certain core of science that is true, and so is knowledge. But then, the problem still applies for most of science. While we think that a certain law counts as knowledge, really only the structure underlying that law does. Other types of realist might argue that science at least gets approximate truth, but this requires that we weaken (2) such that knowledge only requires approximate truth.

How should we get out of this problem? And can you point to anyone who has explicitly wrestled with this issue?