I am the author of that software.
The filter offsets are computed from the linear fits of the position vs. temperature for each filter. See this thread for examples: SGP AF Logviewer v1.2 Release
Now, for this to work, you need a large temperature difference so that there is enough shift in position for a reliable linear fit of the data (for each filter). Typically, one would run an offset sequence within SGP for a large part of the night, maybe a night of bad seeing or with the full moon.
In your case, there is no temperature shift to speak of, so you cannot use that filter offset measurement I would say, for any of the filters. The fitted curve for the Ha shows the problem:
As for the Blue filter specifically, the code returns NaN because the three points have exactly the same temperature, so the linear fit didn’t work.
If you want to get filter offsets from this dataset, then I suggest a simpler method. Simply compute the mean position for each filter using the list, then calculate the difference between filters.
Mikael