Why aren't messages being scanned for spam?

There are several things to check if it appears that spam is not being caught by PureMessage:

Was the message routed through the PureMessage server?
The default PureMessage policy adds an "X-PMX-Version" header to mail from external hosts. If the message does not have this header, it was not processed by PureMessage.
Is F=T set in sendmail's INPUT_MAIL_FILTER line?
The "F=" parameter in the sendmail.mc configuration file determines the action that sendmail takes if the concurrency_limit specified in the pmx.conf configuration file is exceeded or if the PureMessage milter is otherwise unavailable. If this value is not set to "T", messages will bypass PureMessage when the milter is down or busy. See "Configuring Sendmail" in Installing PureMessage for UNIX for more information.
Is the PureMessage Policy configured to deliver messages before they are scanned for spam?
Delivery actions can be assigned to any policy rule, regardless of the rule's position in the policy script. If a message matches a rule with a delivery action before being processed by a rule that scans for spam, the message is delivered without scanning.