1) Because it isn't always worth the cost
2) Because we don't always fully understand the systems we are modifying
Which hearkens back to, "Don't fix it if it ain't broke". Inefficient is not automatically broke.
Don't fix it if it ain't broke seems to run counter to innovation. I'd try for improvement on a functional system if there's a positive expected outcome.
Heh. Forget IANAL; we need IANAA.