Remembering that there's no possible way we could know how many years away a thing is; the numbers are clearly a guesstimate.
They're more of a rounded (hence the clustering at or away from given numbers) measure of the remaining 'effort' versus 'difficulty'. Very likely there is some absolute time required (experiments take time to run), but throwing more people (and more importantly, more resources to run experiments and gather data) at the problem /would/ get to an end result faster (though how much faster is open to debate; it's sort of like asking how much effort does it take to win a top 30th percentile prize from a lottery).
