If your nanna smokes and lives to 100, it doesn't prove that smoking has no impact on your health.
For every smoking nanna with a long life, there are another 50 that die prematurely and hence the conclusion that smoking is bad for your health. It is called the availability bias.
Fyfe is a steal at pick 20, but the data shows that for every Fyfe picked around that range, there are on average 50% of players that make no impact to the AFL. For a pick in the top 5-10 that percentage drops off markedly and the chance of impact is greater. That is why the points are skewed.
The problem is it is a completely imperfect science. There are so many variables and probably the biggest one being how do you rank the quality of a player. For example, how would you compare (say) a Stephen Hill vs Daniel Rich vs Steel Sidebottom vs Luke Shuey. Games played? Disposals? Kicking efficiency? A qualitative assessment? How much more "valuable" is say Nic Nat (taken at 3) vs Phil Davis (taken at 10) in the same draft? Or Zac Clarke, taken at 37 who has probably played ~150 ordinary games vs Steven Motlop?