That is kinda bad practice for many reasons (significant increase in production time, significant rework issues, significant database bloat, though it likely makes wall joins a bit easier…), so you may want to re-consider this approach holistically across the workflow. There are other more efficient ways to get material takeoffs which don’t require individual modeling efforts.
In this specific instance you’ll get only partial results with my method (and likely @Andreas_Dieckmann’s method as well), meaning that as an automation solution it will kinda fail. This is because only one element can be the room bounding one, so everything from the interior room to the exterior face of the wall won’t be selected. You also can’t just select walls with no rooms associated as they will exist inside and outside the building. If you are ok with the partial results and having to manually check what was tagged as exterior or not, test out my method; but If you want all of the mini walls which make up the exterior walls, you’ll likely have to come up with something else.
I think that likely selecting all walls by type (all WRB, all brick, all limestone, all gypsum sheathing, etc) will get you most of the way there, and you can then create a hull of the interior rooms, and do a polycurve containment test to remove the interior exterior walls (if that makes sense). Lots of geometry calculation will have to occur a result of this (again another drawback to the many walls approach is the significant increase in number of walls to think about). Hopefully your building is a very simple construction.