Parasitic drag.
There's gonna be a breaking point when you're expending more energy to compress mix than you're getting back out of the compressed charge power potential.
I think it works on all engines, but some better than others, and there's a point you fall off the curve.
Also has a lot to do with air flow. How much compression you REALLY have at rpm is debatable. I would be nice to be able to measure dynamic compression.
This is where port geometry intrigues me. You leave a nice sharp acute exhaust roof when you raise it (after cutting the band) and you will get more dynamic compression than a flatter curve.
And then, compression=heat, so you're getting a less dense air charge then a cooler motor.
So it's all balance, I think.