What does the severity of the average incident have to do with whether I use a particular mode of transit?

We could add lots of fender-benders between taxiing planes to push down the average severity of each plane "crash." Then they would still be less deadly per mile travelled, but also less severe per incident on average than other modes of transit.

Hard to find up to date data, but looks like about two orders of magnitude, from ~5 deaths per billion miles to 0.05.

(Miles traveled is somewhat charitable to airlines and assumes trips are fungible between planes and cars.)

I'm not sure how you mean it's "charitable?" Could you explain?

Sure, sorry. Planes cover distance very quickly, and people use them for e.g. a six hour flight covering 4000 miles that they would not otherwise drive; this situation makes planes appear safer "per mile" than alternative valid metrics like "accidents per hour of travel" would.