For what it's worth, GNS theory originated as GDS theory (with "Dramatism" instead of "Narrativism") over on Usenet, specifically on rec.games.frp.advocacy.

The idea behind it was to identify the priorities people use to make specific in-game decisions - GM calls, if you will.

So a Gamist decision is one made for reasons of "fairness", a Dramatist decision is one made in order to make things cool and dramatic, and a Simulationist system is one which is made in order to make the game more "believable" (which is not the same as being "realistic").

Forge GNS is almost-but-not-quite the same, and comes with a heavy dose of value judgment.

Specifically Gamism is about the metagame challenge. It's about testing your OOC ability to do what the game asks you to do. Narrativism is about theme, it's about asking moral questions. Simulationism is about exploration which is loosely defined as "everything else."

So to go back to the OP's "guns" example, it is more accurate to say that in a Gamist game guns may or may not one-shot-kill people, depending on whether you want that to be part of the strategic challenge, in a Narrativist game guns may or may not one shot kill people, depending on whether or not you want that to be part of the moral question, and in a Simulationist game guns may or may not one shot kill people, depending on what you want the rules of your fictional world to be.

GNS is basically a hopeless mess. It doesn't adequately describe the different experiences people want from games, or the different ways in which systems go about providing those experiences.

The Riddle of Steel, for example, is touted by Ron Edwards as a "Narrativist" game, even though it provides detailed and realistic rules for combat (with one-shot kills in point of fact). This is alledgedly because TRoS is designed to "raise the question of what is worth killing and dying for". Some sarcastic people (like, say, me) suggest that in fact Ron defines it as Narrativist because he enjoys playing it.