These plots show the difference at 60000 counts between the flux
obtained from 2nd order polynomial fit of flux vs. exposure time and
the flux predicted by the linear term only, normalized by the polynomial flux.