I advise to set filmic middle grey to 18%, then use exposure to compensate globally the mid-tones, then come back to filmic to set up the white and black relative exposure.
Why ? Because it’s more fool-proof, because it anchors the full dynamic range mapping around mid-tones, which can safely be assumed to contain most of the important details in an image, and because it results in a better-balanced filmic curve (where black exposure ~= -white exposure) which is easier to control.
TL;DR : it makes a better-behaved and safer workflow.
But setting filmic middle-grey from 18% to 9% is exactly the same as adding +1EV in exposure. Except for the modules that come between exposure and filmic (because if you push exposure by 1EV, then your RGB values could go past 100%, which is not a problem for the pipeline, but the parametric masking GUI gives you control only in the [0;100] % range, so you can’t set up a key-framing between 50% and 120%, for example).
In future versions of filmic, the scene-middle grey might be hard-coded to 18.45% and disappear from the GUI, because it’s redundant with exposure, and on second (third ?) thought, it only duplicates features.