I looked at how we are doing it on Cacti and it appears that we count the "drops" counter as "non-conformed", then use the difference between "pre" and "post" policy rates to display another line item which is actually the real "drops", calculated locally with an RRD maths script (the very painful method I mentioned originally). To do this in Observium would obviously be a substantial amount of change to the way in which Observium handles the data for this particular set and so I'm not suggesting for a moment this should be done, just pointing out a way that would result in getting correct data and covering up the total lack of thought that Cisco put into their OID structure in case anyone is interested in getting the real drop and re-marking data for their policies :)
Are you sure? The difference between pre/post always seems to equal drops in all the instances I've looked at.
adam.