Yet champions of science (I include myself in those ranks) continue to believe that scientific inquiry is a source of knowledge about the universe around us. 'Knowledge' has traditionally been defined as 'justified true belief'. If our scientific endeavor isn't even interested in whether what it says is true, then it's hard to imagine how its results could still be called knowledge.
I'll say that generally speaking, I'm a scientific realist. I persist in thinking that most of the central terms of our scientific theories possess
reference to the real world, or at least are intended to refer. In your own chemistry, I think that most chemists believe that atoms and molecules really do literally exist, and aren't just conceptual conveniences whose role in chemical theories allows chemists to better predict outcomes when they mix liquids in glass bottles together or create stinks with their Bunsen burners.
http://www.jstor.org/stable/2708449?seq=1#page_scan_tab_contents
Even the author of the OP seems to have reverted to the realist mode of thinking that he was otherwise dismissing when he wrote about caloric, "We now know there is no "heat fluid" known as caloric", suggesting that while caloric theory still works as a predictive instrument for observations as well as it ever did, science can't help thinking about what does and doesn't exist and caloric theory has seemingly been shown to be wrong about that.
Right, we agree on that.
'Truth' needn't imply logical necessity. If a proposition isn't a tautology, it can still be true.
I'm very much a fallibilist regarding science and everything else humans say they know, in the sense that every statement that we intend as a statement of truth may in fact be mistaken. (I'd even extend that to mathematics and logic.) That doesn't imply that we can't sometimes be right. I'm confident that we often are and that we already know a great deal.
That brings up the problem of the so-called 'pessimistic induction', the observation that if all of our outmoded scientific beliefs now seem to be false, it's only reasonable to think that from the point of view of the future everything science currently believes will be recognized as being false too. So we don't really know anything. I think that's the problem that the author of the OP is addressing, if only implicitly.
https://en.wikipedia.org/wiki/Pessimistic_induction
I kind of follow Popper in thinking in terms of verisimilitude (truthlikeness). While outmoded theories aren't entirely true, if they work at all then they do succeed in capturing something about reality. And the author is probably right in thinking that one way to capture that idea of truthlikeness is in terms of empirical adaquacy. Geocentric cosmology was and still continues to be a very good engine for predicting the future positions of heavenly bodies in the sky. Similarly, Einstein's theories will continue still work as well as they work today, even after they are eventually superceded by other theories in the future. They will continue to retain whatever verisimilitude they already possess.
But... even though we can say that medieval geocentric cosmology will works today as a predictive engine for observations of the sky from the earth, it's harder to argue that it continues to be right about the sun orbiting the earth once a day. That just seems to be wrong.
Old and outmoded theories might continue to produce useful predictions of observables, but the way they go about doing it (whether by geocentric cosmology or imagining the existence of vital spirits, caloric or phlogiston) might turn out to bear little resemblance to what is actually happening in the physical world.