Consider patents. I agree with the idea of patents, but totally disagree with the US implementation. Likewise, I agree with giving discounts to non-profit organisations, but the more I dig the less I agree with the US system for awarding the status.
But that wasn't the argument made in the post. It wasn't "we would discount to non-profits if only there weren't so many lame orgs with that status" it was "why would we ever do this?".
And how would you change that? I mean, you can't actually have the government decide what is "the greater good" and what is not- the whole point of charities is that they are non-governmental entities; they work on problems the government is either not addressing or addressing ineffectively.
btw: I think you're looking for 'a priori'
Also, I think you mean a priori.
Do you have strong opinions about what is a bad way to help poor people, perhaps even harming them? There's a non-profit doing that.
Do you have strong political opinions? There's a non-profit, probably a very well-funded non-profit, promoting the opposite with things you consider lies.
That's before we even get into corruption or just plain administrative competence. Specific examples would just set the thread aflame; it's pretty obvious "hideously ineffective" and "[what] I find full-stop evil" non-profits exist. Your list may be different than my list, but we could both come up with one. Unless you have no strong opinions, in which case we could probably still find something in the political category to get you worked up if we tried.
My only issue with his comment is that he's challenged an argument on the basis of a faulty assumption (that charities do good) by introducing a new assumption (that charities might do good, but often do evil) without defending it. So from the standpoint of defensible logic it isn't much better.
I see another commentor made a point about the International Tobacco Growers Association. That's all I was looking for... something we can sink our teeth into that allows the skeptics among us to reject Conclusion A for Conclusion B.