One approach to the question of which objects discussed by a science are real is to designate as real all and only those entities posited by the most explanatorily powerful theory of the science.
Conclusion
The approach (of saying anything posited by the most powerful theory of science is real, and anything not posited by that theory is not real) is flawed.
Evidence
Most theories posit entities only on theoretical grounds.
Evaluate
Saying that something is posited "solely on theoretical grounds" means that it's not attempting to describe something that actually exists. We might posit a perfect circle, a perfect vacuum, frictionless space, the Higgs Boson, etc., and we don't do that in order to convey "We think this is real!"
We do that because it's useful for some reason in terms of the internal logic of the theory. Einstein posed a "cosmological constant" that was basically just his way of fixing the math so it matched reality, but he hated it, since he didn't believe it was a real thing.
If we were predicting a classic "If PREM, then CONC" type answer, then it would sound something like: .
However, we might also notice that this approach is only concerned with "the most explanatorily powerful theory". The objection is based on what is true of "most scientific theories", but this objection is only relevant if this trait of "most theories" actually applies to "the most explanatorily powerful theory".
Goal
Look to connect "treats theoretical entities as real" with "flawed approach / shouldn't do it this way", or look to establish that "the most powerful theory" is similar to "most theories", in the sense that it, too, posits entities solely for theoretical reasons.