DCC - Postfix MTA mail hub implementation question

We run an ISP  with tens of thousands of users and thousands of domains. 
 Hundreds of domains
are sent email when no valid email address exists.  These are excellent 
candidates for creating
spam traps.

Our current configuration is similar to the IMgate anti spam 
configuration described at:

    http://imgate.meiway.com/IMGate-dual.cfm

We have 4 Postfix spam filters running as a mail gateway (IMGate in the 
diagram).  These feed a
set of mail delivery servers.  Our current configuration stops a lot of 
spam, but not nearly enough.

Our goal is to insert DCC into the Postfix system (IMGate in diagram), 
preferably issuing

    550 - rejects

on DCC detected "many" spam messages.  The "550 - reject" is an effort 
to discourage
spammers from continuing to send to our domains. I think this may not be 
possible though.
Other messages I would like to flag as:

    "Subject: [Probable SPAM] $Subject"

for lower probability hits.  

In reading the documentation and the reference at:

    http://www.rhyolite.com/pipermail/dcc/2001/000098.html

it appears that DCC needs to be inserted into the mail delivery machine. 
 This is _very_ undesirable
for us.  

Have I read the documentation wrong?  Can DCC be implemented into a mail 
hub / relay environment
where local delivery does not exist .. preferably with postfix.

Before someone asks, we ran sendmail for about 6 years and have been 
exclusively postfix for the past year.
We see about 4x performance with postfix over sendmail.  We are _much_ 
happier with postfix.

Thanks,

-Steve Camp