Nothing -> something that works (in any way possible, spaghetti code, written quickly) -> something that works (clean refactored code that can be maintained/improved/changed).
You can't skip step 2. If you do, you will likely spend more time overall.
Who skips step 2 without spending a lot more time on a non trivial project? How did you measure this skipped time in real world cases? And what makes it justifiable or not? Where do you get that information that this person(s) does?
Basically; almost all IT projects are delivered too late and for too much money (including games) (and usually they are still crap code); spending any more time on skipping 2 is clearly not justifiable.
I don't think anyone in a real life scenario can actually get this done; I know plenty of (open source) projects where step 2 is skipped, but not in time realistic for any commercial project. Note that I don't think step 2 is a bad thing to do; prototyping something and rewriting/refactoring is not a bad thing as long as you don't ship it as the actual product.
Nothing -> something that works (in any way possible, spaghetti code, written quickly) -> something that works (clean refactored code that can be maintained/improved/changed).
You can't skip step 2. If you do, you will likely spend more time overall.