those numbers need some tweaking. For example, the big fear is random gun violence for which one can rely only on the protection of their own weapon. Well, based upon the below (from your source),
...hence that 17k homicide figure has to be reduced by somewhere between 52%-85%. Moreover, not all homicides are gun-related (and clearly, there is some overlap between gun-related and knowing people). The wiki article was, shockingly, very unclear about the gun-related portion of the 17k homicides. Regardless, lets say its more like 2.5k -8.8k (likely to be on the lower end of the range) for random gun-related homicides, lets say more like 1:123,000. I grant, still more likely than lightning, sure, but 1 in 123,000??? I still call that a pinhole risk. YMMV.
Moreover the operating assumption is that gun ownership renders one immune to being one of those statistics. It may reduce the likelihood, I grant you, but by how much? 10%? 50%? 99%.
I'm not anti-gun. However, if someone said, hey brother, you have a less than .001% chance of being the victim of a homicide, you better gun up bro!, i would probably say, "I'll sleep fine tonight."