Steals is not the only defensive statistic. You are completely ignoring all actual defensive stats
“It's not just the steals or Boston's defensive rating when Irving is on the court, it's his obvious effort and desire to contribute on that end of the floor.
In his six seasons in Cleveland, Irving held an average defensive rating of 106.9. Through 11 games with Boston, that number is down 11.4 points to 95.5, even lower than Boston's league-best rating of 95.9.”
“Kyrie Irving is putting up some of the best individual numbers of his career this season. Here's a look at his career defensive rating via NBA data, as well as his points allowed per possession via Synergy Sports data.” http://www.espn.com/nba/story/_/id/21329484/kyrie-irving-boston-celtics-flat-dominating-defense
Umm.... I don't know where you are getting your defensive rating numbers, but per nbawowy.com, so far this season Kyrie has been on the floor for 767 defensive possessions during which the opponents have scored 802 points which is a defensive rating of 104.5 points per 100 possessions.
With Kyrie OFF the floor, the team has surrendered 511 points on 509 defensive possessions for a defensive rating of 100.5 points per 100 possessions.
If you are using the individual Defensive Rating numbers from basketball-reference.com, keep in mind that those are estimates, not measurements. And they give Aron Baynes the same 111 offensive rating as they do to Kyrie, when in fact the team has actually only scored 103.5 points per 100 offensive possessions with Aron on the floor. In other words, I would take those numbers with a big grain of salt.
Take 5 minutes and compare the data between NBAwowy and NBA.com, you will find a lot of difference in the statistical data. NBA.com is the official site of the NBA. Quoting these off sites leads to a lot of misinformation and confusion.
I've spent a lot more than 5 minutes on all these sites (Yes, I'm a nerd).
It's not as simple as asserting "NBA.com is the official site". That doesn't really make any of it's derived numbers more accurate. What matters is how a value is arrived at. If they could show that their offensive/defensive rating numbers were calculated based on measured points per a measured number of possessions, that would be great. But so far, it's a bit of mystery just how their numbers are calculated. At least I haven't been able to find their published formulae.
They already have some known weirdness in the area of possession counting already because the 'pace' numbers for their boxscores are fractional (i.e., "92.45"), which makes no sense. Possession counts are integral.
Without an explanation of _how_ they are getting their possession count numbers, given the weird pace numbers it's unfortunately suspiciously likely that they are estimating based on event counts such as FGA & TOV stats, but that is prone to over counting possessions (because ORBs can result in multiple FGA on a single possession and there is no way to correct for that unless you know how the ORBs are distributed across possessions).
Evan Zamir (the statistician who created nbawowy.com) calculates his possession counts the correct way, by examining the play-by-play game-log JSON data (which he gets FROM NBA.com) and basically tracking possession changes and tracking player presence on the floor. It's not 100% infallible, due to some shortcomings in the data, but it's pretty close and the numbers are at least presented discretely so you can see what the inputs and formula are for derived values.
I'm not going to claim nbawowy.com is infallible by any stretch. Their USG calculations tend to always come up noticeably short of 100% on 5-man lineups. Of course, that particular number is an estimate and it's not a big deal.
Getting possession counts exactly right is tricky. There are several tripping points. For one, an offensive rebound normally does not start a new possession -- it simply continues the current one. However, if it is a 'team rebound' due to say, the defense tipping it out of bounds there is a chance that a substitution could occur in the middle of the possession. Same with things like a kicked ball. Similarly, when two free throws are taken, players of often subbed in between. Technically, from a TEAM and GAME perspective, this is still just the same possession because there has been no official "change of possession". But different sites handle that differently since the players on the floor have changed. Some count it as a new possession only if players have changed. Others just go ahead and give more players credit for participating in the one possession (this is probably the correct way).
And as I noted, some sites don't do actual possession tracking but rather simply estimate the number of possessions during a period of time based on other stats.