Estimation of fish stock size distributions from survey data requires knowledge about gear selectivity. However, selectivity models rest on assumptions that seldom are analyzed. Departures from these can lead to misinterpretations and biased management recommendations. Here, we use survey data on great Arctic char (Salvelinus umbla) to analyze how correcting for entanglement of fish and nonisometric growth might improve estimates of selectivity curves, and subsequently estimates of size distribution and age-specific mortality. Initial selectivity curves, using the entire data set, were wide and asymmetric, with poor model fits. Removing potentially nonmeshed fish had the greatest positive effect on model fit, resulting in much narrower and less asymmetric selection curves, while attempting to take nonisometric growth into account, by using girth rather than length, improved model fit but not as much. Using simulations we show that correcting for both entanglement and size selectivity produces accurate estimates of mortality rates, while correcting for size selectivity only does not. Our study demonstrates an approach that increases the accuracy of estimates of fish size distributions and mortality rates from survey data.