They're not all mistakes. I'd even say most of the time they are an example of E2 working exactly as it should. When I'm writing a writeup, I should link to anything that could be relevant to someone reading the node--whether or not the node I'm linking to actually exists... yet. And that's what's key. The node I'm linking to doesn't yet exist, and maybe I'm not even planning to write it up myself, but someday someone will. The simplest example is mentioning the name of a famous person. There are plenty of major celebrities who have yet to be enshrined in a node on E2, and yet it's a safe bet that sooner or later someone will get around to it. So if I mention one of them, certainly I should link to the node that will someday be created. The same holds true for famous concepts; it's just easier to see with people, since they have such distinctive names.
Take a look at Using the Web like E2 for musings on how to do the same in the web at large.