Here are results for the most recent codes, similar to the Nov 19th results, in at least taking the same effects into account, but with new changes with the intent of removing a z bias which lead to a distortion bias and luminosity dependence. 

From this point on, the left side is the Nov 19th result, and the right side is current results.
 


Besides less luminosity dependence, a previous issue in the distribution seems to have been resolved (except for the highest luminosity run):

As for the howfit comparison:
Generalized model: [0]/(Radius*Radius) + [1]/Radius + [2]
Original Howfit:  [0]*((3191./(Radius*Radius)) + (122.5/Radius) - 0.395 ) / 15823 

An interesting observation is my early October results (before these final calibrations) had a closer relationship between the models, though I had accidentally included the extra constant.  Though our fits have never showed much improvement via the inclusion of the constant (or a physical one at that), it is interesting to note the difference (below is the October result, not the fit for the most recent data):