Evolutionary conservation scores, quantifying the degree to which sequences have been constrained by selective pressures, provide a means to draw biological inferences on the importance of proteins and their constituent residues. Because of this, these scores have historically underpinned the prediction of missense mutation pathogenicity. Despite their sensible informational contribution, the robustness of tools heavily relying on conservation have proven to be somewhat limited. Recognizing this shortfall in pathogenicity prediction accuracy, the missense tolerance ratio (MTR) was developed. By leveraging the ever-expanding set of available human sequencing data, these scores employ population variant counts and structural information to evaluate the extent of missense mutation assimilation at protein residue sites. Interestingly, regions identified as intolerant using these scores were significantly enriched in pathogenic variants. Moreover, the integration of MTR scores within a machine learning model yielded a highly accurate pathogenicity predictor. While traditional conservation measures observe evolutionary constraints over substantial spans of time and across a diverse range of organisms, MTR scores focus specifically on the modern human population. By normalising the scores and comparing them to one another we have explored the intersection of these approaches, and found evidence as to why conservation scores may be insufficiently informative to make pathogenicity predictions. Traditional conservation measures designate the majority of the human proteome as highly conserved. In contrast, evaluating missense mutation tolerance through population data reveals that the majority of residues in the proteome possess a high degree of tolerance. This substantial disparity between the two approaches implies two key points. Firstly, in relying on conservation scores for pathogenicity prediction, crucial insights into fundamental, human-specific biology are likely being overlooked. Secondly, there is a potential synergy in combining both approaches, suggesting that their amalgamation could enhance pathogenicity prediction beyond the capability of either method in isolation.