Recent developments have cast doubt on the value for the electron stopping power for graphite used in dosimetry. Information has come from a number of areas: a new measurement of the mean excitation energy, I, for graphite; experiments indicating that the crystallite density for graphite, rather than its bulk density, should be used to evaluate the stopping power; experiments in medium-energy x-rays comparing free-air and cavity ionization chambers and Monte Carlo calculations of the behaviour of graphite-walled cavity chambers in ⁶⁰Co gamma-rays and medium-energy x-rays. The stopping powers for graphite currently recommended were evaluated using an I-value of 78 eV. The use of the more recent value I = 86.9 eV reduces the graphite stopping power for ⁶⁰Co by around 1.5%, which in a simple analysis increases the W-value for air by a similar amount. However, such an increase is in conflict with the evidence that the W-value is close to the value 33.97 eV currently recommended. This paper presents the various components and constraints, analyses the uncertainties and shows that the use of a graphite I-value of 82.5 eV, a W-value of 33.97 eV and the crystallite density of graphite reasonably explain the experimental observations. The standard uncertainty of this I-value determination is 1.5 eV.