That's about the same comparison I had, except around the female voice region of the on-wall M22 v2. It didn't bother me at first, but a visitor pointed it out to me one day that she didn't like the way it sounded (too bright), so then I did a lot of a/b testing with various material between on-wall and bookshelf and the bookshelf also had a smoother response in the upper regions as well.
Then with some later testing with a RC sound meter, I discovered that the on-wall had comb filtering problems, whereas the bookshelf did not. While playing a constant tone, I would slowly walk the sound meter across the room. With the on-wall, it would quickly fluctuate up and down. With the bookshelf, it remained steady.
I attribute these abnormalities possibly to the proximity of the speakers to the wall. With the wall interacting with the sound. I don't know how this might affect the in-cabinet or in-wall units.
It's possible some of these problems have been corrected with the v3 units.
