[kokanee]

We can argue about what the article did and didn't imply, but what's interesting to me about the issue you raise is that among lurkers there is probably a much lower rate of fake/spam activity, since there are fewer reasons for a bot to log in and not tweet. Couple that with the fact that lurkers are generally the vast majority of users on any platform, and that alone could explain the discrepancy between Twitter's 5% number and SparkToro's 20%.

[pseudo0]

Services that sell followers and spammers "aging" accounts generally would look like lurkers. Twitter could probably get an accurate estimate with the amount of analytics they have for internal use only, but of course they might be incentivized to not try very hard.