A long time ago I noticed that the sponsored links are less useful than the real links. Searching for a Brand model xxxyyyz, the sponsored link triggers off Brand and is generic for the catalog of that site. The real links, possibly into the same store, will be for the xxxyyyz page.

I've also run into sites that put every keyword for the site in the header of every page, so the search engine can't find the right page. I understand that SE's don't even look at the header keywords anymore.

Generalizing it, if the search engine sees the same footer on every page, it can treat that as spam noise.

I think microformats are of interest here. You can indicate that this page is built using X rather than about X.

I've also run into the problem where every dictionary/glossary page fills a search because both words are on the page. They are not connected though; it just lists all those words separately.

Maybe "nearness" is a good indicator that content is really relevant to both search terms at once. I think Google doesn't support "near" anymore though.