Saturday's New York Times contained a story quoting my opinion on the Goodmail debacle. (I was wearing my Ferris Research hat, natch.) It also had a pie chart, attributed to Ferris Research, illustrating the proportions of legitimate email that are sent to businesses, sent to consumers, and sent by spammers. The caption of the chart implied that 20% of email gets accidentally deleted or quarantined by spam filters. Ouch. While "false positives" are still a significant problem, this figure is of course far too high. Unfortunately, the sense of the original statistic seems to have been lost in the editing process.
Typical false positive rates experienced by spam filter users are closer to 0.1%. State-of-the-art filters can achieve 0.001% -- equivalent to about one legitimate message per month.
The figure that I gave the NYT was the "lost" proportion of legitimate, bulk email -- e.g. legitimate direct marketing and transactional messages. This is roughly 20%, but dropping fast as better spam filters are implemented. While the Times' caption wasn't wrong, it was apparently misleading without the original context, as illustrated by the requests for clarification I've since received!