Question about 3-phase simulation for an adjustable shading system

Dear all,
First, I am using the “3-phase simulation method for a vertically adjustable shading system” by Sarith Subramaniam for my model, which divides the whole window aperture into several groups, performs a separate simulation for each group, and adds the results together( As the picture below shows)

But I found that my simulation results are always lower than expected values. Then, I wonder if this approach is valid or not.

Then, I have run a test to investigate this method a little bit, and my testing is as below.

  1. Case1: In this case, I will divide my windows into four groups, and use the same method as the above picture to calculate the illuminance for a series of selected points. The blinds are assumed to be pulled up in this case, and all these four sub-windows are clear.
  2. Case2: In this case, instead of dividing the windows into four parts, the windows will be modeled as a whole in this case. Also, the blinds are assumed to be pulled up.
    ( you can find all my model files in this shared drive. 3ph_testing - Google Drive)

Result, ideally if the method for adjustable blinds is credible, it should give us the same results as case2 which takes the window as a whole, and the scatter plot of the illuminance values for all points between these two cases should be a straight line. However, as the picture below shows there is a significant discrepancy… So can anyone help me check if there is something wrong with my simulation, or if there is some issue with this method?

It looks like there is enough spread in your results that you could simply be seeing differences due to Monte Carlo sampling variance. You might try running rcontrib (or rfluxmtx) with greater -ad settings and smaller -lw settings. This is just a guess, as I have not looked at your model.