I don't think that there is anything inherently less reliable about projectile based weapons. When you play against other players online, you're not actually seeing their character on your screen, but rather a "dummy" character that mimics their actions as closely as possible. The clients and server send packets back and forth to keep everything in sync, and there are some fancy algorithms to compensate for lag and stuff like that. When it comes to raycasts and collision detection, this is all done on this "dummy" character locally on the client, and then the client reports a hit to the server, and so unless you've got some really bad teleporting going on, projectile based weapons should be just as reliable and consistent as hitscan. If you learn how the two actually work, you'll realize that they both work in a very similar manner. I think that when projectile based weapons seem inconsistent, it's either because of poor netcode in relation to client/server communication, where the client fails to report a successful hit to the server, or that it's simply perceived as inconsistent due to lack of good visual feedback, just like Overdoziz mentioned earlier. If you watch this video, you'll see the shield flash when he shoots the enemy, which means that the hit detection is actually working just fine on the local machine, but is failing to notify the other machines in the network that a successful hit has occurred.
Ah memories. I would never have remembered that map on my own. That lag I definitely recall. It sucks when your view of the game is just to too different from what the host box sees. When you shoot and shoot and shoot people while they miss (as far as what you see) and then you just drop dead. That clip didn't show much of the short-distance teleporting since there wasn't any good strafing footage. In those cases, from previous Halos, it never mattered that our boxes saw us light up someone's shields cuz our boxes didn't have a say in the hits if it wasn't the host. It would have been great to have the game starts show our hit % from both the host and from our box side by side so that we could see who got hosed because of the network conditions. I gotta say that the new server hosted Xbox live has been a big step up from what I'm used to. Often forces me to admit it when I just suck out loud. I find it hard to believe that the Xbox One marketing team doesn't use this as a prominent sales pitch.
If you say that lag is not considered a factor in hit registration then this would be true. But in practical experience I don't know how this would be true. Hitscan may be impacted by lag if the server and your client don't agree. But the practical impact on projectile hit registration is compounded by the lag also affecting your lead. Hitscan can simply make that lead fully independent of lag. That is my thought anyway. This is why I have been saying that hitscan emulating lead would be the best of both. You have to lead by the distance and lag has no impact on how much lead is required. Well... I haven't gotten to a POC yet so I cannot say I can prove my point.