I have heard this 80% figure cited before, on a quick scan cannot find a reference in Kilcullen's book nor on Google. My recollection is that it came from a RAND study.
These quotes can be alarmingly misreading without understanding the full context of the study that determined this. Some possible variations of the study may be:

- counterinsurgents supported substantially by a foreign power, or all counterinsurgencies regardless of context?

- what time frame did the study address (frequently these studies addressed cold war insurgencies)

- the hard question, what does won mean? Did the enemy capitualate? Did they quit fighting? Did the fighting reside for a few years only to flair up again?

- of the 80% defeated and the 20% of the insurgents who won, what were their characterisitics? Were they politically based (communism), identity based (religion, race, etc.), etc.?

Just for the heck of it:

-how many times have insurgents lost in Afghanistan?

-assuming that there really is such a thing as population centric and enemy centric strategies, did anyone win using a population centric strategy?

Perhaps the 80% figure is true, but I generally find these findings to be too generic to be of any use in the real world. They are to often used to justify unwise policy. It is probably true that a 200lb guy will kick the crap out of a 120lb guy 80% of the time if they engage in hand to hand combat, but if you're the 200lb guy getting your butt kicked by a 120lb guy, that bit of trivia isn't overly useful.