The only thing I can think of is that the cam doesn't have the processing power to encode all the pixels in the sensor, so you can either crop the frame (which reduces the field of view) or you can use the whole sensor and subsample (skip pixels/rows like video DSLRs do), which reduces the resolution but maintains the full field of view.
edit: looking more closely at the specs, the HD hero takes 5MP stills so it must be a case of variable subsampling rates. Given this I can't see why they couldn't produce a 1080p signal by subsampling the whole sensor. I wonder if it's something to do with the lens, i.e. it's too soft in the corners for full HD.