Exploratory engineering does put firm limits on future capabilities. There are many possible steam engine designs, and I'm sure those we use today were unimagined in the 1820's. Yet none are more efficient than an ideal Carnot steam engine. The same principle applies here.

Yes, and for aneutronic fusion reactions the maximum drive efficiency is really high. Bremsstrahlung x-rays are still a problem in terms of heat dissipation but shouldn't cut into efficiency too much.

EDIT: Well, I should say that the propellant mass/delta-v is really high and that most of the fusion energy will can end up as kinetic energy of the propellant and ship. But the vast majority of that kinetic energy will be the propellant rather than the ship. To maximize ship kinetic energy over fuel energy you want the propellant to exit the ship at the same velocity the ship is traveling which is very far from what you see in an Epstein drive. But fusion provides tons of energy so that isn't going to be a figure of merit we care about.

True, I didn't consider aneutronic fusion reactions. I don't think the sources I'm thinking of do, either.

Here's one source that does: http://toughsf.blogspot.com/2019/10/the-expanses-epstein-dri...

I think I was underestimating how much energy could be lost to x-rays on some of those. D2-He3, the aneutronic fusion everyone likes to talk about is still over 80% efficient but H1-B looks really bad if you're not driving a heat engine with it.

Yes, though the H1-B reaction itself emits no electromagnetic radiation at all. The x-rays are from bremsstrahlung radiation. Some boron fusion projects think they have ways to avoid that (which is necessary to avoid the plasma cooling too fast to achieve net power).