I think your original plan is better. Being on YouTube doesn't mean only being on YouTube, and forever. People will download your YouTube videos, etc.
Suppose you want to make the best 720x1280 possible from your material. The image quality will be judged by the 95%, not the 5%, of material.
With your second plan you'd throw the 1080x1920 material into a 720x1280 timeline. But FCP6/7 does a not-so-fine reduction. By editing in a 1080x1920 timeline instead, you can throw the whole 1080x1920 output file into Compressor for its best downscaling to 720x1280.
For similar reason, when editing in the 1080x1920 timeline, you wouldn't throw in the 720x1280 material. You'd first put it through Compressor, for best upscaling to 1080x1920.
Dennis Couzin
Berlin, Germany