Nutrient transport across ecosystem boundaries by migratory animals can regulate trophic and biogeochemical dynamics of recipient ecosystems. The magnitude and direction of net nutrient flow between ecosystems is modulated by life history, abundance and biomass, individual behavior, and body element composition of migrating individuals. We tested common assumptions applied to nutrient transport models regarding homeostasis of species' body element composition across space and ontogenetic stage. We quantified whole body phosphorus (P) concentration of three life stages of wild Atlantic salmon (Salmo salar L.) from three distinct populations in Newfoundland, Canada, to evaluate the importance of river of origin and life stage as predictors of salmon %P. We found that life stage was a more important predictor of salmon %P than river of origin, and that %P of post‐spawn adults migrating downstream to the ocean (i.e., kelts) was more similar to %P of juveniles migrating downstream to the ocean (i.e., smolts) than it was to %P of adults migrating upstream to spawn. We then compared nutrient flux for the three rivers over a 20‐year period calculated with body composition values extracted from existing literature and our direct measurements to evaluate how assumptions regarding spatial and ontogenetic homogeneity in salmon %P influenced the observed P fluxes. We demonstrate that assuming equality of kelt %P and adult %P results in an overestimate of net nutrient flux to rivers by Atlantic salmon and the erroneous conclusion that Atlantic salmon populations are unconditional sources of nutrients to their natal watersheds. Instead, Newfoundland's salmon populations are conditional sinks of freshwater P, which is the opposite functional role of Pacific salmon. Our results highlight that a better understanding of intraspecific variation in body element composition of fishes is a prerequisite to determining their role in global biogeochemical cycling.