ACTION_EXPIRED_DATED Stats

I'm trying to assess the effectiveness of having started using
ACTION_EXPIRED_DATED with multiple options.  I started on 2 Nov 2006,
by setting:

	ACTION_EXPIRED_DATED = {
		'default':      'confirm',
		'6M':           'bounce',
		'1Y':           'drop'
	}

In the 105 days since that time, I've had the following stats:

	31500   DROP ACTION_EXPIRED_DATED/1Y
	2716    BOUNCE ACTION_EXPIRED_DATED/6M
	306     CONFIRM ACTION_EXPIRED_DATED/default
	2384	CONFIRM other
	2690	CONFIRM all
	2323	DROP to BOUNCE_ENV_SENDER

In the 105 days previous to that, I had the following stats:

	11683	CONFIRM action_expired_dated
	13206	CONFIRM other
	24889	CONFIRM all
	8926	DROP to BOUNCE_ENV_SENDER

Here are my interpretations of this data:

* The number of BOUNCE_ENV_SENDER drops decreased by roughly 75%.
  I take this to mean that 75% of the emails sent to expired dated
  addresses were to bad sender addresses that were not detected by
  SPF or the MTA.

* I take it for granted that a huge percentage of confirmation
  requests are confirmation requests resulting from a SPAM.
  If that's true, then my ideal goal would be to lower the number
  of CONFIRMS to be exactly equal to the non-spam emails I receive +
  BOUNCE_ENV_SENDER drops.  In that case, the only people that get my
  confirmations are those who are not sending me SPAM.  

  Previously, 36% of my confirmations went to undeliverable
  addresses.  Meaning that 64% of my ocnfirmations went to valid
  addresses.  This suggests that a significant majority of my
  confirmations went somewhere and unnecessarily took up space.
  Now, 86% of my confirmation requests go to undeliverable addresses,
  meaning that only 14% of my confirms go to real addresses.

I don't have a degree in statistics.  So the probability that I'm
making an error seems high to me.  Is there some way I should be
interpreting this data differently?

I frequently post to USENET.  When I do, I use a real email address.
Which means that TMDA is much more likely to be involved in
a joe-job attack than if I posted using a fake email address.
I'm trying to limit that impact by configuring TMDA to handle
USENET address scraping more intelligently.  It seems to me that
I want to send confirmation requests only to those people who are
actually sending me non-spam or to fake addresses.  Am I thinking
about this goal correctly?  Is my logic for achieving it sound?

Thanks in advance for any feedback you might provide.