I am talking about the accumulated error part near the end of the article, so your (uncharitable) assumption is not correct.