Maybe they're incomparable.
i think they are. or at least there is limited value in the comparison. for 3pt shooting, i've seen it referenced that you need a sample size of around 750 attempts before you have more signal than noise in the dataset. that's for a single shooter, for team stats i'd assume you'd need an even larger sample, although maybe not much larger.
for 2pt shooting and turnovers, you're comparing a '16-'17 team that has amassed those stats against an average opposing defense with a 103.8 rating vs a '15-'16 team that faced a 99.0 mean opposing defense (the opposing defenses are actually less different than i thought they'd be). maybe there is more value in these latter comparisons than i initially thought.
the ft shooting numbers, which are by far the best measure of shooting ability, are almost identical.