Optimized 41 area cycles and then spotted a way to save 1 area :(
Area seems to be restricted by the movement of the output atom rather than glyphs and arm space. Furthermore, you need the 3 atom input to reach the middle output.

Most space efficient movement I found was rotation about the 3 atom input (using input supression). This takes a bunch of area, but I use it for gylphs and output (and it leaves a nice critelli spot in the middle).
