This is always going to be a difficult nut to crack as CCB works optimally when predictability of when an agent will become available is high. That, of course, only really happens with a large pool of agents or with low variability in talk-time. What you appear to have is the equivalent of the airport check-in lines where economy and business are served by the same desks and the latter has priority. If you're standing in the economy line you will simply have no way of predicting how many business passengers are going to turn up and get attended ahead of you. With CCB you could consider having a smaller group of agents that will take low priority calls first -- again a bit like having a few check-in desks serving only the economy line. Then, try basing your EWT only on that group of agents activity. Alternatively, this is a case where you should consider using an alternative callback mechanism such as the Agent Request API and only have callbacks made once the agent is available to take the low priority call.
