"What we really want is to minimize the number of infections resulting from any given number of sexual encounters; the flip side of this observation is that it is desirable to maximize the number of (consensual) sexual encounters leading up to any given number of infections."
He's saying that there's this ratio, infections per sexual encounter, that they want to minimize. And the easiest way to minimize this is to increase the denominator. If you have a universe with a skewed (self-selected) distribution, then as you add more people to a category, you'll increasingly get more marginal people. In this case, it's people who have fewer sexual partners and hence are less likely to carry AIDS, resulting in a reduced chance of infection.
Whether this is the metric you care about is debatable. If you're a public health official, you probably care about the absolute number of infections, not the number per sexual encounter, and so this only helps if it brings the probability low enough that the disease can't sustain itself.
Similarly, if you're a married guy who just wants a faithful and devoted wife and kids, you don't care about the chance of infection per sexual encounter, you care about the absolute chance of infection within your life time. This is minimized by not having sex with anyone, or, failing that, having sex with someone who only has sex with you.
Only people who regularly have sex with lots of partners would benefit from this - for them, the metric is chance of infection per sexual encounter (since they'll be having lots of sexual encounters anyway), so why shouldn't they?