Litmus has some very clever code. When they first released their analytics feature, I reverse engineered it because I couldn't believe they could track forwards. Their "time spent reading message" was another cute hack.
Agreed, at best they're useful as a relative measure (version A of the message got more reported forwards/prints/whatever than version B). I doubt the raw numbers are anywhere close to correct. Heck, it's hard enough to track "opens".
But most of their hacks are not 100%