I think the framework basically makes sense. I see potential weaknesses in the evaluation of whether a factor is threatening or non-threatening, as this can be quite subjective.
Perhaps this can be broken down a bit more systematically or a rough decision tree can be developed.
For example something like this (purely schematic, content has no meaning):
Of course, it is not easy to build a framework that covers all scenarios and eliminates any subjective deviation. But maybe you can at least develop something that objectifies and standardizes a large part of it. 80/20 like 