Thanks for the tip, but when I tried to run it under mzscheme I got the error: map: all lists must have same size; arguments were: #<procedure:+> () (1). Also could you explain why set! and list-ref are bad?
You are doing yourself a great service by learning Scheme - that can fundamentally change how you think about programming. Reading the first chapter of SICP[0] will change you forever. Yet right now you're using Scheme as a Python/C/Ruby with a strange syntax, while it's a totally different language with its own idioms. You should learn them to master Scheme.
As for your particular questions:
1. map yields an error under mzscheme
I'm at work, and didn't have access to a proper scheme, so I used an online REPL[1], which apparently is more forgiving. Regardless, you can write your own map (a good exercise!) that stops as soon as one of the arguments is exhausted. Make it tail-recursive[2] as well!
2. list-ref is bad
Lists are beautiful data structures designed for recursive algorithms. If you are using lists and not using recursion (or a hidden recursion in the form of map), you're doing something wrong. list-ref uses list as a vector, which has a performance implications - your algorithm is O(n^3) while mine is O(n^2).
3. set! is bad
Margins are too small for a proper explanation :), but basically set! has side-effects[3], and functional programming should avoid having them.
Using (map + (cdr L) L) doesn't work since they're different sizes (with DrRacket that I used, I see that it worked for you in your other comment, it is a very nice solution!). I'm not sure of an easy constant time way to append a zero to the end of (cdr L). The way I did it is similar to how you did it without map: