I frequently re-write my articles because the "main point" of the conclusions doesn't match the "main point" of the introduction. While writing up the results, my ideas about the purpose of the paper change, so I have to go back and re-structure the initial argument.
Apparently, I'm not the only one to have this problem, based on the post "Always Write the Results First" by Dr. Isis. She recommends that we put together articles in the following order:
1) data figures and tables
2) outline the argument
3) gather supporting citations [I never do this, but I can see the appeal]
4) write results and discussion
5) write methods
6) write intro and abstract
I'm embarrassed this never occurred to me before.