To be fair, the large square can not be cleanly divided by the smaller square(s). Seems obvious to most people, but I didn’t get it at first.

In other words: The size relation of the squares makes this weird solution the most efficient (yet discovered).

Edit: nvm, I am just an idiot.