First off, I don't own one of these fine digital scales,
But, I will comment on general features of some.
A scale with a resolution of 0.001 gram is only slightly better than 0.02 grains.
0.001 gram calculates to 0.0154 grains. A little better than 0.02 grains.
BUT (40 grains= 2592 milligrams)
A 40 grain charge on a milligram scale will have counted 2592 counts,
But in grains mode (0.02 grain resolution) will have counted 2000 counts.
The 100 gram Class
M2 calibration weight that is normally supplied with most scales has a tolerance of +/- 16mg. That's somewhere within 0.016 grams light to 0.016 grams heavy.
That's about +/- a quarter of a grain at the calibration point (+/- 0.247 grains).
A 100 gram
M1 Class calibration weight has a tolerance of +/- 5mg, about 4 counts on a 0.02 grain resolution scale. Again, +/- 4 counts at full scale is likely less at lower loads.
Don't even consider M3 Class weights.
That establishes the full scale end point of an accuracy curve.
With a good zero, measurements at a small fraction of full scale are not effected much by full scale calibration. Zero (auto zero?) can change the calibration curve from the low end.
Zero and full scale set up the end points for accuracy CURVE of the scale
Scale linearity, usually measured from 10% load through 90% load will determine accuracy in most practical applications. Like reloading?
To KNOW the TRUE weight of a load, say 40 grains, you can rely on the new specifications supplied by the manufacturer (if they include linearity in the specs) or you can test NEAR your target weight with a Check Weight.
Look at weight specifications when picking a Check Weight.
Maybe a M1 Class 5 Gram weight (+/- 1.6mg, or +/- 0.025 grains)
is probably good enough to check your charges.
How many COUNTS +/- is the Class of weight you picked?
Now, does the TRUE weight of your charge matter? Or is it just SAMENESS?
If you consider an almost perfect scenario, +/- one half of one count for a good electronic zero, +/- one half of one count for a single reading, NO full scale error, and NO linearity error, the BEST you can expect is
+/- 0.1 PERCENT Accuracy for a 40 grain charge. Probably good enough?
Real world results? Probably NOT that good.
(might want to check my math

)