Concentrations of people with high morbidity rates due to obesity, crime and other issues drag the average down. I think the US can be separated out into two or more distinct groups / societies that will have dramatically different stats. There are cities that are more dangerous than central American countries renowned for their crime, so it's unfair to group them with other areas - the US is a large, diverse country so comparing it with much smaller countries isn't like-for-like.
I'm also always wary of the life-expectancy stats for Japan and they seem to measure it based on people claiming pensions, not that anyone would claim someone was still alive just to get a cheque, but ...