In a New York Times op-ed piece, former CIA employee Joseph Weisberg explains why "classified information is not the same thing as secret information." Weisberg says even widely known information, such as the fact that intelligence agents often pretend to be diplomats (a practice to which Weisberg only alludes, because his former employer would not let him discuss it explicitly), may be classified if confirming it would create operational difficulties. According to Weisberg, agency guidelines allow classification of pretty much anything:
When I worked in the C.I.A.'s directorate of operations (now called the national clandestine service) in the early '90s, we were told that information was classified when it involved sources or methods. It seemed logical that sources were classified. These were actual agents who would be put in jeopardy if their identities were revealed.
But practically everything the C.I.A. does could be considered a "method," so the C.I.A. can decide that almost anything relating to its work is classified. You'd probably want this latitude if you were running an intelligence agency. But one of its unfortunate byproducts is that no one, inside or outside the intelligence community, really knows what classified information is.
Still, Weisberg sees loose classification rules as a good thing on balance, since there may be sound reasons why the CIA declines to confirm public information. It's hard to see how this rationale fits the example with which Weisberg begins: The CIA would not let Valerie Plame specify the dates of her service in her new book, even though the agency itself provided that information in an unclassified letter that was published in the Congressional Record. And there's clear potential for abuse if the CIA classifies anything that might make life harder for the CIA, since that would include keeping secrets to avoid compromising illegal "methods" or simply to avoid embarrassing the agency.