Not sure if this is correct, but I've been seeing the same degredation in performance and I wonder if it's related to just general increases in the complexity of the spam filters. One would imagine that over time, the filters have to be more sophisticated to deal with the ever increasing diversity of spam... For me, it's an indicator that it might be time to look at a SaaS option, if only $boss would allow it.
You are correct, over time the spam filters do become more complex. However, just for clarification, mph is messages per hour that were accepted. For better or for worse, these are messages that were accepted by smtpproxy and were processed by superq. cph is connections per hour and can be rather misleading. It covers everything else and has been the bane of support to explain how the reports get their numbers. The bottom line is that, unless you have an external device checking smtp connectivity, all smtp connections should contain at least one message. The top reasons why a message is not accepted by the smtpproxy are IP reputation, Message reputation, and LDAP. TrustedSource IP and message reputation both require that the message be fully recieved, while ldap can refuse during the smpt conversation but it cant drop the connection until the sending server is done or if it had to many failures.
Putting a rate on mailflow only looks good on paper, it rarely applies to the real world. In this case we can see a max cph of 27352. One thing to keep in mind is that when appliances are being rated they are typically only being rated on "Normal email", which usually falls in the range of 1 - 20kb in size, give or take. In the real world, people love to send prettified word documents, over edited xlxs, picturesque power points, and jpegs of, well, anything - these take a toll on mail systems because they are usually in the range of 1 - 10MB and often occur during peek hours. Combine this with poor network issues (client side) and volume spikes then you can start to see how 27352 can begin taxing a system. As a side note there is grey area here, this is all assuming that each connection holds only one message, however they can hold many, many, many more (I once saw a connection that contained several hundred thousand.). This is why it can be difficult to explain the numbers in the reports.
I can say that the ironmail can handle an incredible amount of "Normal mail" in a lab setting. In my lab I have an EG5500 and using a linux smtp tester (xstress, its wonderfull!!) I maxed out at 600k mph, and that was because my smtpo was backing up - the mail servers couldnt handle the load (exchange on a dell r710 and 2950).
Also, you may want to speak to your Sales Rep about SaaS, it might already be included depending on what you purchased.