Is that a 2 Garmin test? If so:
1. Mad props for getting the bullet through the 18" Garmin trigger window at transonic distances for that rig.
2. More or less, kind of, sort of, the velocity decay for these tests looks like 700 fps. Shooting it down to 1300 fps would be closer to 1800 fps of decay. The result is the uncertainty in the velocity measurement is now contributing 2.5X as much to the BC scatter. The stated accuracy of the Garmin is 0.1% or 3 fps at your velocities. Relying on 2 of them would give an RMS error expectation of 4.5 fps. The problem is I've never seen what they actually mean by "accuracy". Is it the maximum systemic error of an individual unit or 1 SD for random errors? There is no doubt that the Garmin is a few steps ahead of the Labradar on triggering and insensitivity to cosine error, but the Labradar gave trace data and signal to noise ratios for individual shots. We can't set up a batch of test ammunition with exactly the same velocities, and I think it's absurd to compare the Garmin to anything optical or the Labradar, so we're left with testing 2 Garmins against each other. It's on my to do list. Side by side, then check at the extremes of the triggering window, then angle one to see just how insensitive it is to cosine error.
I seriously doubt you're shooting anything with plus or minus 8%. Berger wouldn't let anything like that out the door and the accuracy would be shot by the time the BC opens up that much from barrel wear.
How many rounds on each of those barrels and how long are they? Do you usually test with 10+ shot strings? Are abrasives used to clean the barrel? My N570 fueled 300 Norma went 550 rounds before the BC spread opened up from about 2% to over 4. It was only N570 fueled for the first half of that. it would have been OK for inside a mile for half again that and the groups probably wouldn't have opened up until then.
I saved a 245 Berger AB CDM data sheet from the web that gives 8% BC variation for those bullets unsorted from a fresh barrel. Culling 1 bullet would have cut that in half. Culling 3 to below 2%. If you could figure out which 3 to cull.
Anything with a HP used for ELR will do better with sorting by length. One of the Litz books has the BC sensitivity as 1% per 0.010" of length by way of its effect on meplat diameter. That was for a 30 caliber 200 grain hybrid iirc and it'd need to be scaled. That relationship only had a 50% correlation. My take on that is there are other significant contributors and the biggest is likely the barrel. Particularly the condition, size, and alignment of the throat. I tracked several barrels through their life cycles. The BC dropped and the BC spread opened up in a reasonably predictable way. I estimated BC spread by vertical dispersion on the 2200 yard target. Originally I corrected each shot and applied error bars for the precision. If the strings are long enough, say 10 rounds, and the velocity spread low enough, say 20 fps, just watching the length of the vertical stripe works after calcing it several times. These days, I track the upward velocity migration and use that as an estimate for increasing BC spread. I'm not shooting as much as I used to and don't want to waste 2K days on tired barrels. The resolution of the detailed calcs is maybe 2%. The BC spread fades into the noise below that. 2 -2.5K yards is the sweet spot for the BC spread to pop out of the other noise but before other realities start obscuring things.
Early days, I tested just pointing and while it made the BC better, it left the BC spread. The bullet needs a meplat with sharp edges and a diameter of about 15% of the bore. Near as I can tell, that's one thing the bullets known for low BC variation share. My take on it is it's a sacrifice of some zero yaw BC for greatly reduced BC sensitivity to initial levels of yaw. My guess is the yaw variation is what's opening up as the throat wears. The process that worked the best for me was culling out the truly weird bullets, trimming the rest to a common length, using a "hollow pointing" tool to clean out the trimming swarf, then pointing to leave a 15% of bore meplat. After seating, it shouldn't be obvious that the bullets have ever been touched by human hands, much less stamped by the seating stem. I spend a lot of time detailing seating stems. Low seating pressure requirements help and I lube the boat tail junction/back of the bearing surface with Imperial before seating to further reduce the force on the bullet.
Moving from BC spread to BC truing, the effect of air temperature is often overlooked. At 2K a 10 F error in the air temperature estimate appears very similar to a 1% change in BC. The air temperature isn't likely to change that much during a string, but it's unlikely the average air temperature along the path of the bullet is within 5 degrees of the shooting position. Overcast days after overcast nights are the best for truing. After that, learning to lean on the air temperature used by the solver 5-10 degrees can help first round hits beyond a mile. Adjust up after a clear night, adjust down at noon on a sunny day. As with most of this, the largest thing to avoid is end to end errors. 10 degrees up when truing, 10 degrees low at the match. 2% BC error trued into your solver....
