That's basically what the entire article is about. The (very) short version is it really is possible to predict when things will be done, but if you tell the engineers or negotiate with the engineers (ie. turn it into a deadline), that counterintuitively tends to make the predictions worse instead of better.