If there really is some quantum interaction, then this becomes interesting, but from momentum alone the contributions from the sides of the box always exactly cancel the resultant from the mismatched plates (by definition of a closed box).
If you want to say you get less thrust as velocity increases then you run into trouble with the principle of relativity.
So I'm skeptical too, though I think it's worth continuing experiments.
I hadn't noticed the constant thrust problem you mention, I'm now recursing back through the equations I remember to try and work out why it costs more energy to push something that's already moving fast... wait, force applied was never anything to do with energy to start with, impulse/power were. Brain-fart resolved.
If you draw constant power from the batteries, is it impossible to maintain constant acceleration?
Which means, "a reactionless drive is like a car in space" would be a nonsensical concept without a preferred reference frame.