1) The under-represented, regardless of whatever group or subgroup they belong to or are socially identified with, believe as staunchly in the merit principle as anyone else. Heck, maybe even moreso. Indeed, it is the belief that they have traditionally not been fairly evaluated FOR their merit, that annoys the heck out of them. This is the source of many who might surprise you by their rejection of affirmative action initiatives intended to help people just like them. Sure everybody wants a great job, but EVERYBODY wants to feel like they were hired because they know what they're doing, and have skills. Few people feel comfortable being hired for something other than their skills, especially if their co-workers know it.
2) We tend to define "merit" in a rather cold and narrowly circumscribed way, forgetting how much of organizational functioning is dependent on more than mere KSAs, but on relationships and trust too. As a result, it is exceedingly rare that you will find a hiring manager who is willing to make a hire solely on the basis of a stack of professionally developed tests that are internationally recognized for predicting job performance. Ironically, they may place complete faith in on-line software to choose a perfect life partner for them, sometimes sight unseen. But if they care about how that job is to be done, that manager will insist on interviewing the candidate, and/or want some sort of reference from someone whose judgment they trust. Ultimately, they want to know they can work with that person, and they need to see it with their own eyes in order to trust it; validated tests be damned. And since they often expect that person will be working with others already employed n that organization, they will want to know the new person and existing employees can probably work together. I'm not saying that ought to trump all, merely that this is how people think.
So it should not surprise us in the least that employers turn to what is less often "favoritism" but more often "trust-ism", and that what they trust is what they know, and what they know can unfortunately be divided or at least sorted along demographic lines. That doesn't make it acceptable, but it does make it predictable.
Sometimes there ARE subgroup differences in access to key information. People learn about jobs through informal as well as formal channels, and under-represented groups may be a little less likely to report learning through informal channels than majorities are. And when we look at things like who ends up getting summer student placements in government jobs, we see that kids with parents or other relatives working in government are disproportionately more likely to get those jobs; not because of favoritism or patronage, but simply because they knew about the very existence of the jobs in the first place.
But if our tacit notion of "merit" also includes trust, then the real challenge is one much broader than mere employment alone. It becomes one of fostering greater familiarity and trust between demographic "streams", such that merit in the classic sense can be clearly seen every time it presents itself, regardless of what superficial package it comes in.