Just upload the surfaces you are using in the gh definition, not the whole file. But I am like, 90% certain that your problem is the discontinuity of the initial point topology.
M.
pixel select, with a random reduction (cull component). At a quick glance, you have lower 3 rows at about 90% density, next two rows at about 55%, and top two at 17%.
Gilbert Tesselation with propagation in 2 dirrections X and Y. Crack propagation with an angle of 90°.
http://www.grasshopper3d.com/forum/topics/gilbert-tesselation-2