Welcome to WebmasterWorld Guest from 126.96.36.199
Anyhow, while I found "site:" to be maybe 65% reliable (I'll expand on this) I still find the correlation (between the 'site' number and SERP traffic) to be fairly significant. The issues with 'site' I have noticed so far is
a) 'Under-reporting' - e.g., on a 100 page site, the site operator returns 10 pages, but google sends traffic to all 100.
b) There is a fluke that occurs sporadicaly (let's say once in 10 days) when the operator returns rediculous numbers, e.g., on a 100 page site I get 750 indexed pages. I have several sites and this affects let's say one in five every time it occurs.