For a small number of cases, there is a mismatch of path loss between Windows and Linux.
This appears to be a numeric rounding issue related to the LinearLeastSquaresFit function.

After further digging, this appears to be due to small differences in terrain profile. The
same Python code and terrain data was used to create the terrain files on Windows and Linux,
but different OS results in small differences in terrain profile. The extremely small
differences (max difference approx. 1E-9) can cause > 1 dB difference in pathloss in some
cases due to numeric error in the LinearLeastSquaresFit function.

I attached a couple of examples, I have more available if needed.

Input Files:
case_1_itm_inputs.txt
case_2_itm_inputs.txt

Terrain Files:
case_2_windows_pfl.txt
case_1_linux_pfl.txt
case_1_windows_pfl.txt
case_2_linux_pfl.txt

Output Files:
case_1_itm_command_output_windows_pfl.txt
case_2_itm_command_output_linux_pfl.txt
case_2_itm_command_output_windows_pfl.txt
case_1_itm_command_output_linux_pfl.txt