From that experience I did notice that getting listed on a relatively big one like KillerStartups.com always resulted in an avalanche of automatic listings on smaller directories and blogs - as well as a lot of twitter activity.
From poring over google analytics, and talking to users, I concluded that while a lot of that chatter is automated garbage, it does often end up being read by real people who follow the links and sign up for the app.
All of this is to say that potential web users are scattered across too many discovery vectors to solve the problem by simply building The One True Directory.
What could be done, maybe, is build a sort of central repository to which developers could post their apps in a standardized fashion and then have those listings freely available. That repository could work a bit like Crunchbase, except with less focus on money & people and more on categorization & features. Ideally, app directories would then consume and relay that information to their audiences - editorialized or not.
Such an initiative would require massive participation by app developers, which, of course, is the real crux of the matter. In practical terms, there would be little difference between building a solution now or simply flocking massively to one of the already available directories. Which begs the question: why don't we?