You are of course quite correct.

Just for curiousity, is there anything in the codebase to stop $NORM from being negative? If not, then that's potentially an argument in favour of the arbitrary fixed point being below zero. Although there would certainly be bigger fish to fry if such a thing ever occurred.