You could also interpret this in a different way: The experiment did some good, but at the same time established a ceiling after which doing even more good would run contrary to the experiment's interests.
E.g., if some miracle happened and the city council suddenly had the means to include every homeless person into the program, this would actually be highly problematic for the program, because now they'd be lacking a control group. So by running the program, the council may help some homeless people, but at the same time it has an active interest that the other homeless are not helped.
I think this is the difference in general this article glosses over. It's not about doing a little good vs doing a lot of good, it's about not wanting to do more than a certain amount of good, even if you could.