At a certain definitional level, I can see his point.
Think of comprehensiveness in two parts. First, it allows users to find anything and everything that is relevant to a search query (i.e., a user will miss nothing), as they hone in on whatever they are looking for. Second, it means the ability for users to find things they want to know, but perhaps didn't even know they wanted to know it (i.e., a user will be prompted to refine, even change, her or his needs by the very act of seeing comprehensive search results).
Undoubtedly, overall, Google has been a master of this comprehensiveness. In addition, they have improved sensitivities in the words one uses to find better and more relevant results. You see both clearly in their latest local search release last month. Local search queries bring more relevant, geographically accurate results than ever before. And search on specific and even quirky local needs ("vegetarian pizza" or "wifi coffee spots"), and useful results appear. Add to this a cleaner and more sophisticated mapping capability where every result set is numbered (in order of relevance) and the designated number is placed on a nearby map so you can visually compare where your results are geographically.
Yahoo! certainly values comprehensiveness, but their focus in local hones the user into very specific, systematic results sets - in a word, precision. Search for a restaurant, plumber, or most any local service, and one will receive tighter search results that don't merely send you to the relevant Web sites, but to useful summary pages of what you want to know about that service.
While it might be fun to find anything there is about plumbing in Washington, D.C. on the Internet, in the end what a user really wants is a good, close, and reliable plumber. Yahoo!'s focus, therefore, is heavily based on their existing structured data and merchant relationships, as well as in licensing deals for structured data that are actively in development (as they have already cut with Yellow Pages), to create great results sets with specific information one requires to make a local choice, and the ability to act on that choice.
And they appreciate what I have described previously as "nuance" - that people want more than a series of results. They want a way to understand if the result really is relevant to them. Is it expensive? (Yahoo!'s answer: sort results by price.) Is it nearby? (Yahoo!'s answer: sort by distance.) Is the service any good? (Yahoo!'s answer: solicit self-published user ratings and make those ratings sortable.) You may not be able to find a vegetarian pizza, nor necessarily find every Web posting relevant to your search - but you don't get 15 pages of search results either.
My own take, overall, is that the answer to "what is more important in search, comprehensiveness or precision?" at one level must simply be "both." And yet, I think that there is a significant difference from when one wants to search for a research paper on World War II, and when one searches for an Italian restaurant in D.C.
In the former, you want a pretty wide swathe to ensure that you aren't missing something or you are trying to expand your investigation. In the latter, you want to know what Italian restaurants are out there, where they are, how much they cost, and whether they are right for you. Locally, precision is key.
Now, in this example, the result set of restaurants - the universe of restaurants - had better be comprehensive. If my search results do not yield "Café Milano," a popular Italian restaurant here in Washington, then doubts are raised about the veracity of the search over all. But, to me, this is more about the comprehensiveness of structured data, as opposed to everything and anything that might be somewhat relevant throughout the World Wide Web.
I have suggested in the past that structured data about listings vertical by vertical, with great detail about the result set that helps us refine our decisions, will be paramount in the battle for local eyeballs. Combine this with the most relevant unstructured data (news articles, reviews, other relevant postings) and transaction capabilities (schedule, book, buy) and include a user interface that does not bury us in an ocean of vaguely relevant results - and you have a destination I'd bookmark and use every day.
Technology is forever improving, and structured data vertical by vertical is if not buildable, licensable through almost any relevant local search category (folks have long aggregated rich information on restaurants, retailers, lawyers, dentists, etc. and are eager to license it) Today, we are in search like we were in television in 1952 - it's only the beginning.