I doubt the error rate was 0% with human curators. In fact, the article itself does not compare the error rate of the human curators to algorithm, just examines

I doubt the error rate was 0% with human curators. In fact, the article itself does not compare the error rate of the human curators to algorithm, just examines that it was above zero after the humans were removed. I think an acceptable statement to make would be that the human curators probably reduced the number of "fake stories", but that's not why they were removed. It's a different judgement call whether the trending feed is "better" or "worse" in some holistic sense. Furthermore, there seems to be an implicit assumption that Facebook's Trending stories should not contain "fake stories", but is that the intended goal of the service? I actually have no idea, since I've never used it.