WASHINGTON — After seven years of trying, the government thinks it finally has an accurate count of how many secrets it creates: 6.8 million a year.

That’s a 35 percent decrease from the previous count, but nearly all the drop came when the Navy decided it really didn’t originate more secrets than the rest of the government put together.

It’s still just an estimate but it appears that the “top secret,” “secret” and “confidential” stamps were wielded 6,796,501 times in the fiscal year which ended last Sept. 30.

“It appears that this unprecedented drop is primarily the result of more accurate counting, rather than an actual tremendous decrease in classification activity,” said Steven Garfinkel, who has spent the past seven years trying to get the government to produce fewer secrets and count them more accurately.

Garfinkel heads the Information Security Oversight Office, created in 1982 when then-President Reagan rewrote all the rules for classifying documents.

In his annual report to President Bush, released Tuesday, Garfinkel praised the Navy for finally figuring out what he has been trying to count each year.

The government doesn’t actually count its secrecy decisions each year. It samples the number of times some offices decide to classify information. Then it projects those numbers to reflect the number of decisions in all offices.

Moreover, there are two kinds of secrets: original secrets, of which 500,000 were produced last year, and derivative secrets, of which 6.3 million were created. For example, an official creates an original secret by deciding that all photographs from a spy satellite will be kept secret; another official creates a derivative secret each time he stamps one of the photos “top secret.”

A derivative secret also is created when one official copies another official’s original secret into a new document — the government equivalent of the friend who says, “Don’t tell so-and-so, but I just learned that ….”

For 1987 and 1988, the Navy reported creating far more original secrets than the rest of the government combined. As Garfinkel put it in his report to Bush, both the Defense Department and his office “expressed skepticism about the accuracy of these data.”

In 1989, the Navy learned what went wrong. Some of its sample offices were “reporting the number of all classified documents on hand, not just classification decisions that they had originated,” Garfinkel said.