Interesting question. You could further refine it to what percentage is "scientifically" able to be used for an ID to a specific firearm, and what percentage is usable to influence a jury?
I suspect a good portion of the "identification" performed is not to a specific firearm, but to narrow the possibilities down to an already suspected firearm: caliber, brand (e.g., distinctive Glock firing pin), rifling pattern (# lands/grooves, twist rate).
My sister was trying to convince me that at a local trial a "firearms expert" testified that a recovered bullet could tell whether the firearm was fired accidentally or on purpose?!?! Whether true or not, the "ballistic evidence" is used to influence a jury to somebody's benefit and another's detriment.
"The ultimate authority ... resides in the people alone. ... The advantage of being armed, which the Americans possess over the people of almost every other nation ... forms a barrier against the enterprises of ambition."
- James Madison