I think they're saying that the tool may have been used like this:

1) Create an ad targeting straight couples, and hide it from gay couples. 2) Create an ad targeting gay couples, and hide it from straight couples.

If you query for "is anyone running an ad hidden from gay couples", this would return "yes".

However, my hunch (hope?) is that the prosecution was clever enough to account for this, and are specifically finding ad campaigns which are "unbalanced" (eg; there is no ad targeting group X).