I think they just assumed mortality risk had an inverse linear relationship with the natural log with the amount of exercise.

Yeah, the table just gives linear estimates, but all of them do say more exercise is better, I think.