Several automated techniques for assisting engineers in creating mappings between source code (e.g., classes) and architecture elements (e.g., modules) have been proposed. They are generally assessed via precision and recall metrics. However, these metrics can only be evaluated post-mortem, i.e., once an expert created and validated all the mappings. In practice, given a set of mappings, engineers would like to assess their quality without (effortfully) validating the whole set. In this context, we explore a suite of quality metrics as an architectural fitness function for a ranking of mappings, which helps engineers select a useful list of those mappings. We empirically analyzed the evolution of our metrics in two projects using different mapping techniques.