Kellerman compared intra-family homicides and "dope deals gone bad" homicides--all within residences--with the number of dead burglars.
First off, intra-family homicides generally occur after a buildup over time, with prior police responses to complaints. Typically, the firearm had been legally acquired some lengthy period prior to the homicide.
Next problem is defining the success of burglary deterrence by the number of dead burglars. He didn't take into account the far larger numbers of times that burglars had been scared away by a resident's threat of shooting, nor did he take into account those burglars who were held by an armed resident until the police arrived.
SFAIK, the "taking away" of one's firearm by an intruder is a rather rare event. I've yet to read of it, anywhere...