Collider in RCT Subgroup Analysis

The variable does not change - what we do with it does. In b) we treat it as an effect modifier and these results are observed. Now we have to decide if this analysis is valid and that decision hinges on what the variable is concluded to be and if the specific analysis is accepted as the way to analyze such a variable or not. In b) we assumed effect modifier and stratified - I disagree with both decisions