Revisiting the modelling framework for the unresolved scalar variance