I assume that some (probably small) proportion of this 44% has used their account for direct messages, DMs, but these can't be detected? Also, how are protected accounts factoring into these stats? I couldn't find a link to the data itself and the article didn't make it clear how these two factors are being handled.