Defensive win shares is an extremely flawed statistic. It assumes all teammates are equally good (per minute) at forcing non-steal turnovers and non-block misses, and assumes that all teammates face the same number of total possessions per minute. This leads to it over-valuing guys that get a decent number of steals but are obviously not good defensively (James Harden). It also undervalues elite rim protectors that allow a low FG% at the rim, but don't have the block numbers to back it up.
I LOOOOVE advanced analytics but defense is tough to quantify.