I've given this a lot of thought. My chrono data collection has been EXTENSIVE of late. I, too, have seen some disconcerting Extreme Spreads (ES) and done more than my fair share of head scratching.
Like the OP, my initial thought was, "I'm throwing powder inconsistently and/or my scale sucks." But after experimenting with different scales I've reached the conclusion that good powder throw habits (like always tapping after dispensation to ensure everything comes out) make my Lee Auto Drum pretty consistent. In fact, it'll throw more consistently than my scale will weigh it - so I've learned to STOP "chasing throw weights" with constant adjustments to the LAD while loading a batch.
But that still doesn't account for some of the differences. Take this example of the same bullet fired with the same charge weight of powder through the same 2 guns...only parameter varied was the primer (and maybe the COAL but that doesn't much matter for a cartridge with all the space of a .38 Special):
COAL__________Trail Boss Powder____125-gr Lead RNFP bullet____________________Barrel
1.440-1.448_____3.7_______________775______________________112.1_REM 1.5 SPP___5.5
1.440-1.448_____3.7_______________974______________________91.4__REM 1.5 SPP___20
Thus, CLEARLY the FED-100 primer gives tighter ES values. That has to be a factor.
Then I hypothesized that, "Maybe each gun's throat or cylinder gap caused the ES to be measured from it to be tighter or looser." But some thorough review showed that to be wrong, as this .44 Mag load data shows:
COAL__________Lil' gun Powder______240-gr X-Treme CPFP bullet________________________Barrel
See how the 17-gr load is tightest in the 4" gun but loosest in the 6.5" gun? Then see how the tightest for the 6.5" is the 19.6-gr load which is one of the looser data sets for the 4" gun. I have other data that seems to indicate one gun is always tighter in ES than another...and then I collect new data that refutes that assertion.
So what does all this mean? A lot, and very little, all at the same time.
It means there are a LOT of parameters that go into ES. Furthermore, it shows that ES is just a thumbnail and NOT statistically significant. So what's it good for? Getting an IDEA for a decent load...nothing more.
After you get an IDEA then you need to put a statistically significant number of rounds over the chrono and see how the standard deviation (SD) pans out. I'm talking over 20 data points here.
I'm about to start doing just that, with .500 S&W Magnum rounds no less.
Thus spoke Zarathustra.
Footsoldier in the Conservative Insurrection of the GOP.
Remember, only you can prevent big government!