A comprehensive understanding of uncertainty sources in experimental measurements is required to develop robust thermochemical models for use in industrial applications. Due to the complexity of the combustion process in gas turbine engines, simpler flames are generally used to study fundamental combustion properties and measure concentrations of important species to validate and improve modelling. Stable, laminar flames have increasingly been used to study nitrogen oxide (NOx) formation in lean-to-rich compositions in low-to-high pressures to assess model predictions and improve accuracy to help develop future low-emissions systems. They allow for non-intrusive diagnostics to measure sub-ppm concentrations of pollutant molecules, as well as important precursors, and provide well-defined boundary conditions to directly compare experiments with simulations. The uncertainties of experimentally-measured boundary conditions and the inherent kinetic uncertainties in the nitrogen chemistry are propagated through one-dimensional stagnation flame simulations to quantify the relative importance of the two sources and estimate their impact on predictions. Measurements in lean, stoichiometric, and rich methane-air flames are used to investigate the production pathways active in those conditions. Various spectral expansions are used to develop surrogate models with different levels of accuracy to perform the uncertainty analysis for 15 important reactions in the nitrogen chemistry and the 6 boundary conditions (ϕ, Tin, uin, du/dzin, Tsurf, P) simultaneously. After estimating the individual parametric contributions, the uncertainty of the boundary conditions are shown to have a relatively small impact on the prediction of NOx compared to kinetic uncertainties in these laboratory experiments. These results show that properly calibrated laminar flame experiments can, not only provide validation targets for modelling, but also accurate indirect measurements that can later be used to infer individual kinetic rates to improve thermochemical models.