• Old Hayes calibration was off by as much as 10%!!!
• Atmosphere is most likely culprit

Table 1: Instrument Science Report NICMOS 2005-002 (STScI)

Figure 10: Instrument Science Report NICMOS 2006-002 (STScI)
Ratio of the merged G096+G141 fluxes to predicted flux for the five stars with pure hydrogen NLTE models. The average and rms of the ratio is written on each panel. The brighter primary standards in the bottom three panels define the absolute sensitivities; and the average of these three ratios should be unity at any wavelength.