Fake accounts are culled? Apparently not very well.

To take one simple example, half of Bieber's followers are fake. I'd be willing to bet that between 1/3 and 1/2 of all followers on public accounts are fake.

http://www.bbc.co.uk/newsbeat/22104058

Then on top of this, I'd like to see how many Twitter users post a few Tweets and then give up. It's not hard to calculate that data just by sampling a large number of users from the API. I'd argue it's an extreme portion of the supposedly active user base.

Mostly Twitter is shallow (as in limited use and interaction) consumption platform, with relatively few highly active producers.