Let me preface this by saying that I'm not falling on either side here. I look at things as a system. What parts of a system require absolute precision, or if some parts allow variance, how much variance is acceptable while still meeting whatever specification or goal you have set?
To that end, I just got out my RCBS Uniflow powder measure purchased in 1976 to run some stick IMR4831 (i.e. not the newer short cut variety) through it to see how much charge weights vary. I did this same test many years ago to see how much technique mattered but that data was lost somewhere.
The measure isn't mounted on a stand, it's mounted on my press using that flat bar they supply with it held in place with whatever die is handy. Also no baffle, it's filled to about an inch from the top and the cap replaced.
Two techniques were tried, one good bump at the top and one at the bottom of the stroke and a double bump at the top of the stroke - mine fills at the top stroke. It was set with an RCBS 505 scale to a "close enough" nominal weight but for this trial I was more interested in the total variance than any absolute. Normally I'd set the measure by throwing three charges and getting the total weight, i.e. for a 40 grain weight I'd set the measure when it threw an exact 120 grains.
The formula (total variance / nominal charge weight) * 100 was used to get total variance as a percent of the nominal charge weight.
Mathman can chastise me on my experimental parameters if he wants, I didn't measure moisture content of the powder or anything like that.
Just wanted to see what would happen throwing charges in a "normal" manner.
First try, nominal 40 grains of powder, one bump at top and bottom of stroke:
39.5, 39.5, 39.4, 39.6, 39.6, 39.6 - 0.2 grains total variance, 0.5% of nominal weight
40 grains of powder, double bump at top of stroke:
39.9, 39.9, 39.8, 40.0, 39.8, 40.0 - 0.2 grains total variance, 0.5% of nominal weight
Then I set it to a 60 grain charge to see if a bigger charge weight produces a bigger variance. In this case it seems to have done so. I only used a single bump at top and bottom for this trial.
60.2, 60.4, 60.0, 60.1, 60.4, 60.0 - 0.4 grain total variance, 0.67% of nominal weight
Again, not falling on either side of the argument, it's a hobby, do whatever you like to do. Is .5% variance in powder charge going to have a significant or insignificant effect on grouping? If absolute precision is the goal then one can get 0.0 grain variance by throwing a light charge and trickling up. The only intent here is to give some empirical data on variance of a worst case powder just by throwing from a measure.
Btw, if someone wants to repeat this experiment with a measure mounted more solidly and/or using a baffle, please, have at it.