With gun or knife, the probability of a kill depends upon the motivation of the person using the weapon, and the levels of strength, skill, preparedness, and armament of the assailant and the target. I would expect that there would be some situations where the kill probability would be less than 1% with either weapon, and others where it would be greater than 99%. Of course, such kill probabilities could also be achieved (though in a narrower range of situations) by an assailant armed with nothing but hands and feet.
If an assailant is going to attack someone until the person is dead, and if neither the person nor anyone else is going to stop the assailant, then odds are pretty good the target is going to die. If the assailant's only weapon is a gun and the assailant lacks the strength to attack via other means, then if the assailant runs out of ammo the target might survive. Likewise if the assailant's only weapon is a knife which breaks. Otherwise, the 'single-attack kill probability' won't matter much since the assailant can simply attack as many times as needed.
Incidentally, another factor confounding measurement of kill probability is the difficulty of determining an attacker's motivation. If an knife-wielding attacker is trying to injure but not kill an attacker, the probability of a kill may be quite low (depending upon the injuries inflicted). A knife is probably more capable of inflicting predictable non-fatal injures than a gun. If the government will believe a claim by a knife-wielding assailant that he wasn't trying to kill his target, that may cause knives to appear less effective than they are.
Ok, all things being equal then.