I think that methodology is wrong. If they analyzed only one instance, then the whole follwer-followee thing might be completely off. And while they measured how many replied frequently, it lacks the “how many replies they got” side

It’s like comparing engagement on whole network vs via only one edge