I'm impressed by extremely difficult challenge here. I wonder what C-Munk, our specialist of impossible tasks would think of it. Sure with no horizontals at all, many ambiguities resist and some glyphs need a hard brain effort for read but i think it would be really difficult to go closer than you did with such a minimalistic concept.

though there's only one glyph i really don't get : it's the "s" made like the "c" but functionning much less

Neurone error, you have no idea how much it means to me to be referred to as a "specialist of impossible tasks".

In the capitals I see that the M, O and W are excactly the same. You could make a "hole" in the O and take a little off the bottom of the middle line of the W. That should solve it.

The lower case has a few of the same as well. You could let the s decend a bit - like a greek final sigma - to differentiate it from the c.

The o, v and u are also the same. Maybe you could do like the n, but upside down and mirrored for v and u.

Maybe make the z a backwards c so it doesn't look like an x.

Oh, and I just realise now that the g, q and y are copies as well. I can't find an easy solution to that right away, but it should be doable.

But there are a lot of good solutions as well. I particularly like the Q, R, S, X, Y, Z, a, e, x.