I'd go with higher cubes because those look squashed. Ratio between height of the cube and it's width (diagonal) should be as close to cos(30deg)/(sin(30deg)*sqrt(2))/2 as possible. So since single cell cube is 25px wide, it should be around 15px high (not 13px as in this system).
16px looks even better but it would clash with "grid lines" creating line clusters.
Having cubes slightly higher so that wall-lines are slightly offset (15px instead of 13px, so +2px, a single micro-grid cell) from the floor-lines makes it better looking and easier to tell apart where walls occlude floors on the picture (or the other way around) or where different levels of floors are next to each other.
4X4 cube should theoretically have 59px height (not 49px as in this system) but it might be more pragmatic to use 57px so that you can stack 4x1 cubes to get to 4-cube height.
Apart from that, this system great! I especially like the rule about vertical lines on right pixels only.
The price I pay for better proportions is that height is divisible by 2 and 7, while floors are divisible by 2, 3 and 6.
But it might make sense that walls have different patterns of division than floors. And dividing walls evenly might be not that useful (except for levels of buildings) and placing things equidistantly between the ceiling and floor on the walls is also not very useful.
16px looks even better but it would clash with "grid lines" creating line clusters.
Having cubes slightly higher so that wall-lines are slightly offset (15px instead of 13px, so +2px, a single micro-grid cell) from the floor-lines makes it better looking and easier to tell apart where walls occlude floors on the picture (or the other way around) or where different levels of floors are next to each other.
4X4 cube should theoretically have 59px height (not 49px as in this system) but it might be more pragmatic to use 57px so that you can stack 4x1 cubes to get to 4-cube height.
Apart from that, this system great! I especially like the rule about vertical lines on right pixels only.