Its normalized by distance, so the fact that 75% of your driving is short trips doesnt matter. You could argue that you need to look at only sustainable trips, but I think its unlikely to change the conclusion that per mile flying is safer than driving on a per mile basis.

You'd need to set some minimum trip distance for which flying is a viable alternative, like say 300 or 500 miles. Then throw out all the driving data for under that. Now, how would you expect this to shift the deaths/mile of driving?

In order to reduce it, driving short trips would have to be more dangerous per mile than driving long trips. I think this is unlikely to be true. Long trips often take place on highways at higher speeds and thus accidents have higher consequences. Drivers are more likely to be fatigued on longer trips. Drivers are less likely to be familiar with the roads on longer trips.

You've probably also got to throw out all flights for which driving is not a possibility. Like transcontinental trips. That may make flying appear safer since flights over oceans are more dangerous since there arent as many emergency landing opportunities.