After: 100 users, 99 accurate retweets.
The study followed 2000 and I can only assume they all retweeted multiple times, like 10, so you get 20000 retweets. Seems plausible to me.