When your looking at on/off court numbers it grabs a lot of factors over which an individual has very little control (i.e. performance of teammates, what opposition is on the floor, etc.) so you can't just assume he's the problem.

But in this case it's not +/- (although I think I remember that Knight has generally fared pretty poorly in the advanced plus-minus stats as well) so much as team offensive efficiency, it's just the points his teams have scored divided by number of possessions the team has had with him on/off the court. Once you get a start to get a large enough sample size it does raise some questions.

If I'm a team looking at giving Knight an 8 figure per year extension and making him the PG of the near future I'd want to dig into the tape and try to figure out why, throughout his career, his teams score significantly better when he's not on the court.