Actually, thinking through it again fresh, I think it does all hinge on what the percentage is - but some other variables come into play (it's not a single percentage fits all) and then it quickly turns to madness because of the complexity of the system.

If the percentage is higher than the unconstrained optimal margin then the cap has no effect. There's no new pressure introduced yet.

If the percentage is lower than the unconstrained optimal margin then the only incentive is to increase the cost to raise the cap until right at the point demand decreases enough that any more cost would actually result in less total revenue. Because medical care is often very inelastic, that'd could quickly be a lot of cost inflation even for just a few percentage point constraint off the optimal margin. This is the part you're highlighting, and that makes sense.

The main counteracting force to this would be that a single insurer does not (theoretically, at least) set the cost of care directly on their own, they (theoretically, at least) compete with each other to negotiate the best care rates to have the most consumers go through them. There are several things which practically get in the way of that though, like how often you can actually change insurance plans or how competitive the open insurance market is (if you even have multiple options, some states only have a single marketplace option) vs just sticking with whatever your work offers.

Between all of that it is where comes back to the common refrains of "and that's why we need to go to a single payer system without profit as the main goal" and "and that's why we need to get rid of the ACA and let the market handle optimal profit naturally". Everybody can't seem to agree which way to go, just that they don't like the current way. Ironically, these approaches effectively map to the 0% cap (single payer, no profit focus) or the 100% (no ACA cap, free insurance market) interpretation options I originally listed.

I'm sure there about a billion other nuances not covered or thought about in this... but at least the comment parses now!