After more research and discussion here and elsewhere I have concluded that:
film: the characteristic curve maps logH (illuminance on the film to the resulting log10(Density) i.e. log10(1/(reflectance or transmittance))
digital: DPR's test curves map the 21 Stouffer steps to 8-bit sRGB.
Therefore both X-axes are logarithmic: the one base 10; the other base 2.
BUT the Kodak/Stouffer wedges have a much greater range 0.05-3.05 (10 EV) of density than film say 0.2-2.3 (7 EV) and none of them have 18% gray at the logarithmic mean. Furthermore, most Zone step diagrams don't have 18% gray in the middle, instead showing 127 or 128 out of 255! (everyone and his dog including CIPA/ISO knows that 18% gray maps to 118/255).
So the Kodak/Stouffer wedges just serve as a reference range for determining the characteristic curve for films - which do vary quite a lot.
And so-called "middle gray" is picked out of the air by various entities i.e. it is not always 18%.
Case in point: most Zone tutorials have Zone V at 127/255 i.e. close enough to 0.5 which, working backward to the sensor via 2.2 gamma, gives about 22% exposure, not 18%.
Recently found this Kodak illustration which maps density to both reflectance and 8-bit RGB (2.2 gamma, also 1.8 gamma)
Unlike other step-wedges this one has intervals of 0.1 density i.e. neither full nor half steps, but about 1/3 step.