The disparity between the increases of 'peak' and 'average' publishers is curious. It may be simply a function of reporting methods. It used to be explained that 'peak' includes ones who report several months at one time to catch up, so August (at the end of the service year) was the highest month, as people were asked to complete all their reports before Sept 1. Now with texts, electronic communication, etc it may be that the effect of the 'peak' is less pronounced. In either case, the 'average' publisher is probably a better measure to compare across years.