- Many 20% projects are not things that would not be easily monetizable: open source work, non-technical things, improvements to infrastructure.
- If the idea is very speculative, exploring it while still getting paid a handsome salary is a much easier step than quitting your job.
- Some things are easier to do if you have all the Google infrastructure to build on: for example the Transit Maps thing below is a lot easier if you can plug into Maps! (But some things would be easier to launch externally.)
- Part of the attraction, like for open source projects, is that it's something different from your main job, so you learn and stay fresh. If you turn it into your main job you lose that.
The theory is that if you do work on a new product and it works well, eventually it will be staffed full-time and you'll be rewarded. Apparently that did happen with Google Now. I doubt it happens every time, but then not every worthwhile startup succeeds.