"We also were told that over 90% should stick of the DSL users"
No, we weren't. In fact, I would venture to say that at the time that Dean made a <bi>similar statement in the CC, you were one of the posters who said that his statement didn't have a whole lot of meaning, as he didn't provide the current sample size for which the retention rate applied, for instance.
He said something to the effect that the uptake rate on contacts with subscribers as of that point, was around the 90% (or something close to that) level. He did not say that 90% should stick.