The problem with the AC graphic should be only measurement/equipament error. The curve varys with the flow squared (constant*Q^2), and simply there's no way for that to happen... probably if you get more data the graphic will show as it is expected to be.