There are no known bugs when your shot fly through the unit but not detected as hit.
Actually, there is, and when real accuracy is off, and extender is on, I simultaneously see the cases when supposed 100% shots miss and projectiles fly through the sprite, not hitting the voxel itself. I just don't have savegame, because I never thought I will address these issues. Nevertheless, your mod makes this particular shot actually hit, thus the problem is solved.
As for the balance - if there's indeed no clue of how close-range shots affected by extender accuracy, may we talk with the designer of this? You see, it's kind of magic and fits game very well.
I'll bring you the case:
You have auto-fire sniper rifle, which has reduced hitchances for any tile closer than 10. At 10 tiles away it gives you 100% vanilla hitchance (which magically won't hit 100 out of 100 shots, but that's cool, too), but 0% at 2 tiles away.
Still, due to extender accuracy, you will likely hit at least 1 time in a queue of 3 shots, but also it may be 2 times or even 3.
I can only guess, that it somehow creates cone of possible trajectories, and the more precise the shooter, the narrower the cone is.
It's not like it will shoot somewhere 45 degrees from a target, right? Soldier with accuracy 120 should shot with little deviation from LoS (line of sight), but still can miss because target is too far away and it's actual contact surface is too small.
The question: what actual mechanics are, both in your mod and extender?
Is it cone mechanics, or just pure formula-based pre-hits (mean, system rolls 1 hit out of 10 and then forcefully pulls 9 hits to the random tiles around the target?)
And, one suggestion for your accuracy mod:
If it checks several points of shooting before deciding which fits best, shouldn't this also affect accuracy? You see, if straight/ out-of-the-waist shot is considered worst, then the obstacle is on the way. If then weapon shifted to a side, such action must have penalty of decreased accuracy. My suggestion is 1/4 decrease of the final value.
Same to melee. I've noticed from the game that if target stays in front of the door, it sometimes can be attacked from adjacent tiles, and vice versa. These hits shouldn't be of same accuracy to the straight attack, because target has natural cover and has lesser area to evade/parry from. I guess, final chance attack success rate should be, at least, halved.
I would also suggest that shifting the hand should affect the chances, if not complicated to implement.
