It doesn't matter because the growth of fog with increasing density is probably small (if it exists at all). But I don't agree this effect (again, if it even exists) is taken into account when the curve is plotted. If the plotted curve represents gross density, then fog is included everywhere. No problem there. But characteristic curves are very commonly plotted showing net density, where a constant fb+fog density is subtracted from all the measured densities. People calculate contrast, etc. this way.