The Google "site:website.com keyword" search:
A) takes the same* (*but vastly expanded) search results for "keyword" but essentially filters out all results that come from other domains by focusing the search only on "website.com"
In other words, which pages should rank based on running the normal algorithm for "keyword" as if "website.com" were the only site in Google's index.
This being distinctly different than:
B) shows which pages on "website.com" contained "keyword" in their source code when Google last crawled
It has always been my understanding that the answer was A, which would explain the search results I'm seeing, but I'm seeing other things claiming the answer is B.
Which is it?