Successfully predicting the effects of amino acid substitutions on protein function and stability remains challenging. Recent efforts to improve computational models have incluzded training and validation on high-throughput experimental datasets, such as those generated by deep mutational scanning (DMS) approaches. However, DMS signals typically conflate a substitution's effects on protein function with those on in vivo protein abundance; this limits the resolution of mechanistic insights that can be gleaned from DMS data. Distinguishing functional changes from abundance-related effects is particularly important for substitutions that exhibit intermediate outcomes (e.g., partial loss-of-function), which are difficult to predict. Here, we explored changes in in vivo abundance for substitutions at representative positions in the SARS-CoV-2 Main Protease (Mpro). For this study, we used previously published DMS results to identify "rheostat" positions, which are defined by having substitutions that sample a broad range of intermediate outcomes. We generated 10 substitutions at each of six positions and separately measured effects on function and abundance. Results revealed an ~45-fold range of change for abundance, demonstrating that it can make significant contributions to DMS outcomes. Moreover, the six tested positions showed diverse substitution sensitivities for function and abundance. Some positions influenced only one parameter. Others exhibited rheostatic effects on both parameters, which to our knowledge, provides the first example of such behavior. Since effects on function and abundance may arise through different biophysical bases, these results underscore the need for datasets that independently measure these parameters in order to build predictors with enhanced mechanistic insights.