Possibly also due to hydrogen spillover effects as the atoms dissociated at the active catalyst sites would migrate on the carbon layers formed. The possible positive role of suitable carbon layers (even deposited from hydrocarbons during reaction conditions, not just the catalyst support) has been highlighted in the catalyst literature too.
His early studies with thermionic emitters were of desorption of alkali atoms as clusters (Rydberg matter) from graphite and graphite-covered metal surfaces. No Rydberg clustering would occur from pure metals without a graphite layer. From the previously linked review:
Isn't the graphite doped with potassium at 7%?