Anyways, in the way that sakky has defined "meritocratic admissions" - that it is contingent in the students who actually apply and decide to choose a school with "meritocratic admissions" over schools that don't follow such admissions policies - that sort of admissions is contingent upon the environment of the country and the attitudes held by its top students, and is completely independent of how the institution would best serve itself - by adopting environmental cues and choosing the best route it can for its own goals.
Let's face it - many universities in other countries practice "meritocratic admissions" too. In such countries, "merit" is often defined as scores on particular standardized tests. Many of these universities are different from Caltech - in that they're larger and often must bear the stamp of "I have the nation's best students period." And with that stamp comes expectations.
Caltech has an unique privilege. It's that national attitudes in a first-world nation aren't overly fixed on math/science, as attitudes in developing nations that want to become first-world often are. And with the attitude that America has, well, Caltech truly does get many of the students who are the most interested in math/science. Many of the top students do go to other universities, of course. Some people are more likely to find math/science enriching if they have another activity to pursue (there is some psychological research to support this assertion - Dean Simonton's books are an excellent place to start). Caltech is not perceived as having such activities (even though it has many of them, just that students may not have as much time to pursue them). It loses students this way. So what? Caltech still manages to achieve its goals, and provide an ideal environment for the students who want that sort of environment. Many of the top students are of different personality types. At least they have the option to pursue environments in concordance to their personality types here - in that they have many options to go through - many such environments to choose from, if they want to pursue scientific research.
Or so this is in an ideal society. Some of them don't get into the top schools - due to admissions policies with too few slots/affirmative action/the school's desire to branch out - take your pick. At least they can still go into state schools. Only that they are so-narrow minded as to actually follow the curriculum in the state school and not educate themselves - and don't have the peers to motivate them in such schools - and that by the time they actually get into graduate school, their fluid intelligence has actually declined (in math/science, there are few initial novel contributions after age 30 - yes, even in the times when 18 year olds could actually contribute something novel. Incidental memory peaks in ages 13-15 (SCIAM Mind April 2007, article on teenage angst in Western societies). Here's an interesting take -
http://home.twcny.rr.com/hiemstra/tlchap2.html ). You could blame them for not motivating themselves - their willpower and what not - only that most people really aren't all that good in motivating themselves and that a nation built entirely on self-study may not be the nation with the most scientific output.
Eh, life's not perfect.