Cultivar mixtures slow polycyclic epidemics but may also affect the evolution of pathogen populations by diversifying the selection pressures exerted by their plant hosts at field scale. We compared the dynamics of natural populations of the fungal pathogen Zymoseptoria tritici in pure stands and in three binary mixtures of wheat cultivars (one susceptible cultivar and one cultivar carrying the recently broken-down Stb16q gene) over two annual field epidemics. We combined analyses of population "size" based on disease severity, and of population "composition" based on changes in the frequency of virulence against Stb16q in seedling assays with more than 3000 strains. Disease reductions were observed in mixtures late in the epidemic, at the whole-canopy scale and on both cultivars, suggesting the existence of a reciprocal protective effect. The three cultivar proportions in the mixtures (0.25, 0.5, and 0.75) modulated the decrease in (a) the size of the pathogen population relative to the two pure stands, (b) the size of the virulent subpopulation, and (c) the frequency of virulence relative to the pure stand of the cultivar carrying Stb16q. Our findings suggest that optimal proportions may differ slightly between the three indicators considered. We argue potential trade-offs that should be taken into account when deploying a resistance gene in cultivar mixtures: between the dual objectives "efficacy" and "durability," and between the "size" and "frequency" of the virulent subpopulation. Based on current knowledge, it remains unclear whether virulent subpopulation size or frequency has the largest influence on interepidemic virulence transmission.