Single-cell expression profiling is a rich resource of cellular heterogeneity. While profiling every sample under study would be advantageous, it is time-consuming and costly. We introduced Cell Population Mapping (CPM), a deconvolution algorithm in which the wide spectrum of immune cell types and states is inferred from the bulk transcriptome based on reference single-cell profiles. CPM was applied to investigate individual variation in lung cells during in-vivo influenza virus infection across a large number of murine strains. The analysis revealed that the relationship between immune cell abundance and clinical symptoms is a cell-state-specific property that varies gradually along the temporal trajectory of cell activation states. We show that this observation can be explained by a mathematical model, in which clinical outcomes relate to cell-state dynamics along the activation process. Our analyses demonstrate the utility of CPM as an efficient cell-mapping tool, and highlight the importance of such a tool for understanding phenotypic diversity.