"Most striking, however, is that when you hold educational attainment and other observable worker characteristics constant, pay is starkly different depending on the firm where you work, even within narrowly-defined education categories, industries and occupations. This is prima-facie evidence that the human capital model in a competitive labor market is an increasingly poor way to explain earnings inequality. Wages for similar workers do not, in fact, equilibrate across firms."
An analogy I like here is one of a basketball school. Imagine you take x students, all of a relatively reasonable capability, and put them into a school. And you give them all an identical education in basketball. Nobody in the world would ever then expect these students to be of roughly equal ability after they graduate the school 4 years later. There would be vast differences between their abilities at that point. Even if you control for obvious genetic factors such as height and perhaps athleticism, there'd still be absolutely tremendous differences. There have been thousands of professional basketball players -- all in the 0.0001% of society, yet even given that ultra-selectivity Michael Jordan stands head and shoulders above nearly all of them yet. And the same is true in most of everything. If you want to go a purely mental game, we have a series of chess players -- Paul Morphy, Jose Raul Capablanca, Bobby Fischer, Gary Kasparov, and now Magnus Carlsen that in turn stand leaps and bounds ahead of a player pool (of their respective times) that is/was once again already made up only of the most capable 0.0001%.
If it's somehow not clear, the point of this is that just because A and B go to the same school and end up with roughly the same grades - that says absolutely nothing whatsoever about the ability of A or B. I would argue that not only is the stated observation not "prima-facie evidence that the human capital model in a competitive labor market is an increasingly poor way to explain earnings inequality" but rather the exact opposite! If companies were recruiting exclusively based upon observable characteristics it would mean these less measurable differences that separate the good (or even the rather less than good) from the great would not be being factored in. That would be rather strong evidence that effective competition was not correlating to positive results. But we have the exact opposite!