After a couple of days of struggling to come up with a simple algorithm involving addition, subtraction and one table lookup, the algorithm for one rotor is complete. In comes a 9, out comes a 3. Even better, the steps for the group of instructions at the bottom are the same as the top, just feeding it a fixed input of 0, done with a call to a subroutine, then subjecting the output of another round of (if negative, add 10). Need to investigate byte savings of making a (if negative, add 10) subroutine...