His hypothesis had nothing to do with the angle of the leaves but rather with the Fibonacci sequence. For some reason he also decided to angle the leaves in his test, which I suspect were the true cause of the efficiency gain.