Google will list any page it finds a link to, whether or not it is allowed (by robots.txt) to fetch and analyze that page. If the page is disallowed in robots.txt, it just lists it as a URL, with no title or description.
The solution to this problem is non-intuitive: You must *allow* the pages to be fetched in robots.txt, and then use the on-page html <meta name="robots" content="noindex"> tag to tell them to ignore the page.
Ask Jeeves/Teoma does the same thing, and as of this month, Yahoo's Slurp is now apparently doing it, too. Yahoo adds the interesting twist of using the link text it finds on the link as the title for the listing.