The only issue is whether the weight error is symmetrically distributed? If a bullet is a couple of grains light, the weight error itself isn't enough to matter at pistol ranges, but if that couple of grains is all missing from just under one spot near the surface on the side of the bullet, its spin will make it wobble in flight. That is, its path will describe a helix around the trajectory it would have had if it were balanced. The helix is OK if it isn't too big, but if it's big enough it can make a bullet flip in flight and keyhole at the target.
Unfortunately, you'd need to make a bullet spinner (dynamic balance tester) or a horizontal torsion balance to measure CG to figure out if you have a problem with a bullet in advance of firing (Harold Vaughn's book, Rifle Accuracy Facts describes these devices). Someone also makes a sonic bullet tester, but these types of devices are more geared toward precision rifle match shooting needs. I would make a piece of paper with with bullet weights in .1 grain increments ruled on the bottom and place each bullet above its weight or above the last bullet of its weight. You'll end up with the bullets making up a histogram; a bar graph of bullet columns showing weight distribution. If there is just one central peak tapering off evenly at both sides, the error is pretty much likely to be random and ignorable. You could go to the trouble to calculate a Gaussian bell curve to see how it fits the peaks of the bullet columns, but mainly you just want the visual aid of the graph. You may see a small separate peak on the heavy side. That indicates something was stuck in the block when those were cast. You may see a few bullets obviously too light and separate from the group, and should be suspicious of inclusions with those. You can separate them out and shoot them separately to see how they group? If they do well, ignore the problem in the future. If they do poorly, pick them out and recycle them into the melt in the future.
Below is a simplified example of what I mean using half grains instead of 0.1 grain increments. You might get two peaks near each other because of having two cavities, but you'll get the idea. The heavy clump on the right is when the molds were jammed open. The scattered ones on the light side probably have inclusion or didn't fill the mold well. I would recycle in either case.
