pyflakes best practices?

[Roy Smith]

We've recently started using pyflakes.  The results seem to be similar 
to most tools of this genre.  It found a few real problems.  It 
generated a lot of noise about things which weren't really wrong, but 
were easy to fix (mostly, unused imports), and a few plain old false 
positives which have no easy "fix" (in the sense of, things I can change 
which will make pyflakes STFU).

So, what's the best practice here?  How do people deal with the false 
positives?  Is there some way to annotate the source code to tell 
pyflakes to ignore something?