Agreed with the experiment flaws, plus the article is way too shady to make anything even remotely reliable out of it. It reads awfully like "Child pwns scientists, saves world and his dog"

Still, what if you consider a fixed volume as the constraint in establishing the comparison? For a flat panel assembly, a given volume would constrain the surface quite obviously. If you consider a tree structure one might be able to pack some more panels in the same volume, and Fibonacci might give us some solution about how to maximize panel density in the volume while minimizing shading interference from other panels and supporting infrastructure.