The diagram on page 13 has linear interpolation plotted between the datapoints. That's one step away from fitting an arbitrary polynomial for the points. Don't do it, implying measurements between point releases isn't very sensible.