They might have other teams using it for intelligence rather than case-building. If they sequestered those from the teams building evidence that usage might not be discoverable.
Not secure against end-to-end attacks: Tor does not claim to completely solve end-to-end timing or intersection attacks. Some approaches, such as having users run their own onion routers, may help; see Section 9 for more discussion.
They are repeating it several times in their documentation, too.
It's not really a bug - there is little that can be done here, IMO.
However, depending on how high the garbage stream must be set to ensure that there is never a spike of real communications higher than that, it could easily be too costly for most people.
So if it saturates your connection for an hour for 6 hours randomly spaced throughout a day, it's not immediately apparent if that's because you're using it, or it's a decoy stream. Varying the amount used (and always adding at least a little extra when in use) would also make it harder to detect.
At least, that's how it seems to me. There may be some sort of cunning statistical attacks depending on the implementation, especially if the attackers have the endpoint under physical surveillance (and notice that your presence always matches traffic increases of some level)