Experiments which determine the product of (W/e)air, the average energy deposited per coulomb of charge of one sign released by an electron coming to rest in dry air, and (LDelta/rho)Ca, the Spencer-Attix mean restricted mass collision stopping-power ratio for graphite to air, in a 60Co or 137Cs beam are reanalysed. Correction factors, e.g., to account for gaps about a calorimeter core or perturbations due to a cavity's presence, are calculated using the EGSnrc Monte Carlo code system and these generally decrease the value of (W/e)air(LDelta/rho)Ca for each experiment. Stopping-power ratios are calculated for different choices of density correction and average excitation energy (I-value) for graphite. To calculate an average value (W/e)air(LBIPM/rho)Ca for the BIPM air kerma standard, each experimental result is multiplied by the ratio (LBIPM/rho)Ca/(LDelta/Rho)Ca. While individual values of (LDelta/rho)Ca are sensitive to the I-values and density corrections assumed, this ratio varies by less than 0.1% for different choices. Hence, the product (W/e)air(LBIPM/rho)Ca is relatively insensitive to these choices. The weighted mean of the updated data is (W/e)air(LBIPM/rho)Ca=33.68 J C(-1)+/-0.2%, suggesting that the accepted value of 33.97 J C(-1)+/-0.1% is 0.8% too high. This has implications for primary 60Co air kerma standards worldwide and potentially for the choice of graphite I-value and density correction for the calculation of the graphite stopping power, as well as the value of (W/e)air.