The Cd measurements agree with the calculation to within the level of precision of the method (1-2%).
The Ba measurements are approximately 5% lower than the calculation (on axis).
The Am measurements are a factor of 3 lower than prediction, which is enormous.
For this particular source, the casing is stainless steel, meaning that depending on the orientation of the source, there could have been serious attenuation. Tests done by Jay and Scott revealed that when the Am source was oriented backward, the rate dropped by a factor of 3, just as was seen in the data.
The Co measurements are about 40% higher than the calculation.
One possible explanation that was entertained was that Compton scattering from passive materials contributed extra counts. Since background was removed from the measurement by the mask-weighting and cleaning processes, this Compton scattering would have to have been from the immediate vicinity of the source or directly behind the detectors, both of which would produce counts that would not be background-subtracted. However, fits to the data show that the main photopeak peak itself is much larger than the predicted photopeak. Any compton scattered events would place counts in bins below the main peak, so this explanation is ruled out.
The Cd measurements agree with predictions to within 1% or so.
The Ba measurements are about 6% higher than predicted (on axis).
The Am measurements are about 13% higher than predicted (on axis).
This discrepancy may be explainable in terms of photons that are scattered into the detectors from nearby passive materials. Am is dominated by 60 keV photons. These photons are more susceptible to Compton scattering than those from Cd or Ba.
The Co measurements are about 33% higher than predicted (on axis).
Like the Am measurements, this discrepancy may be explainable in terms of photons that are scattered into the detectors from nearby passive materials. Like Am, Co is dominated by fairly high energy photons (122 keV). These photons are more susceptible to Compton scattering than those of Cd, Ba, or even Am.
Jack compared the total count rate above 20 keV with the count rate due to photons that deposited their full energy in the detectors (using the fits that Goro did using the monte-carlo-generated depth distributions) and found that 30% more total counts are present than full-energy counts. This matches the discrepancy I found very nicely and seems to confirm the suggestion that the discrepancy is due to scattered counts.