Yep, I came here to say the same thing. The author has misremembered / misrepresented Simpson's paradox, which is much stronger than an aggregate hiding a group effect.