Vegetation fires play an important role in global and regional carbon cycles. Due to climate warming and land‐use shifts, fire patterns are changing and fire impacts increasing in many of the world’s regions. Reducing uncertainties in carbon budgeting calculations from fires is therefore fundamental to advance our current understanding and forecasting capabilities. Here we study 20 chamber burns from the FIREX FireLab experiment, which burnt a representative set of North American wildland fuels, to assess: i) differences in carbon emission estimations between the commonly‐used ‘consumed biomass’ approach and the ‘burnt carbon’ approach; ii) pyrogenic carbon (PyC) production rates; and iii) thermal and chemical recalcitrance of the PyC produced, as proxies of its biogeochemical stability. We find that the ‘consumed biomass’ approach leads to overestimation of carbon emissions by 2‐27% (most values between 2‐10%). This accounting error arises largely from not considering PyC production and, even if relatively small, can therefore have important implications for medium‐ and long‐term carbon budgeting. A large fraction (34-100%) of this PyC was contained in the charred fine residue, a post‐fire material frequently overlooked in fire‐carbon research. However, the most recalcitrant PyC was in the form of woody charcoal, with estimated half‐lives for most samples exceeding 1000 years. Combustion efficiency was relatively high in these laboratory burns compared to actual wildland fire conditions, likely leading to lower PyC production rates. We therefore argue that the PyC production values obtained here, and associated overestimation of carbon emissions, should be taken as low‐end estimates for wildland fire conditions.