They vary by area because of the nature of research.
It is much harder in the experimental sciences to "schedule" results. My PhD has both experimental (lab work) and computing aspects and all the computing research took within the expect time (+- 20%). The lab aspects were much more variable. I had one experiment which I expected would take 3 months to finalise that worked first time and I was finished by the afternoon (this doesn't happen that often unfortunately), while other experiments I thought would take a week ended up taking 6 months.