modelDependent is an indication internally that we can pre-populate measure.xml with argument values without having to dynamically generate values for a specific use case. The most common example is a choice list that have hard coded values of "North, South, East, and West" Those are not modelDependant. If you instead you have a choice list that has you loop through all spaces in the model, that is modelDepedant. Either approach is valid,we just have to dynamically calculate values for xml in the second option.
Having a measure where values change from one model to another is fine, but having a measure where the number of arguments that exist change from one model to the next can be problematic for a parametric analysis. If you are just using it internally or with apply measures now then it may be fine. You can't have an argument there and not show it, but there are a few options. One option is to make it a choice list instead of a bool and and for some models just offer "false" instead of "true,false" It will show as an argument but the user won't be able to change it. You can also use an optional argument. Another option is to completely skip the argument in the argument section, it will require some extra code in the run section to either not attempt to get the argument using the same logic, or to gracefully handle having it there or missing.
