For near max loads, I weigh each one no matter what the powder. Even with extruded powder, if the loads are middle or bottom of the data range, I will use the powder measure. I use the RCBS measure for rifle and a Lyman 55 for everything else I don't load on my Dillon SDB. SW
I guess the answer we've come to is: do whatever gives one the most consistent charges in the cartridges - when thrown loads do such, we do that; when weighed loads do such then we weigh then... and over and above that, we pick our powders to allow such.
For plinking and target loads using small charges, we always throw the loads (a uniflow with micrometer)... and when we have to use powders which can't be thrown consistently, then we weight them.
For the light loads (not in pressure, but mass) we simply found that weighing them, beam balance or electronic scale we couldn't get near the same consistency - using the uniflow with an equalized powder column (*) (!important!) and a friendly powder (aa2, wst waap vvn320 w231 etc) we find that our std error is +/- 0.02grs (or given a 4gr charge a +/- .5% error). With the less friendly powders (greendot amslct) then .03gr. The best we could do with a scale was a measured +/- .1gr or a little less.
One test of how consistent the light/small loads are: is to throw/measure ten, combine them, and see if one see's 10x the expected value on the scale. When we weighed small loads and did such we always found there to be skew either heavy or light, ie, 10 "4.0gr" loads would measure 40.7gr or 39.4gr etc. (this has to do with the +/- .1 or .2gr error always in an electronic scale or lack of readability of beams.) The other test is simply to look at the ES and SDs generated - when one is seeing 10 shot SD's under 4fps and ESs in the 8-10fps range something must be going right. W/re full loads vs starting loads - we don't see much difference: if one picks whichever gives the most consistent charges... it should always be used.
Powder choice is also important - and give the huge selection of choices today, there is no reason to work with a powder which is difficult to get consistent loads. As such, unless we simply need to document a particular powder, we tend to avoid the imr's 4198 and slower (ie, most of them); likewise most of the RL's. This is especially true when one can use hodgdon's rendition of most of the imr's and likewise given their "short cut" powders. But over and above that, most of the vvn's given the same "speed" all throw much more accurately/consistently.
hope that adds something to discussion... and do shoot straight,
(*) an equalized powder colunm in two senses: if one simply loads the powder column and starts throwing loads, one will find a slow shift in charge weights... simply because the column becomes more (uniformly) packed as each powder grain migrates toward the bottom - we'll throw 20-30 loads and return them to the column before we bother to calibrate the measure. Likewise, depending on the ambient temperature, we'll allow the powder to come to the ambient. Our powder magazines keep the stored powder at 53-56degF year-round - if the ambient is 95degF (eg) then one will find in a few minutes that the powder on the sides of the column will have a different density vs that in the middle, and having no control over where the grains for any particular charge comes from - one will find their charges varying all over the place, ie, either quickly throw the loads before the heat sets in, or let the whole column come up to temp, and then throw the loads.
The other important point with calibrating a measure: don't adjust the micrometer just before throwing a load, ie, doing such will put the small column just above it into a different packed state than the rest... if one throws 10 loads and finds they're at 40.3gr (ie, 4.03 per charge) and knows that by taking 1 "number"/"stripe" out of the setting one will be at 4.00 per, do such, but then throw 3 or 4 loads (and put them back into the column) before dropping loads into cartridges - those first couple of loads will be heavy compared to what one was expecting. [otherwise, with a graphic chronograph that first shot becomes painfully obvious.]
any guess for which string the cartridge builder tweaked the micrometer adjust and forgot to throw 3 loads before starting into the actual cartridges? (at first we thought this effect was an artifact of a cold barrel... until we shot the 5 PWs in reverse order and still saw it.)
(**) this whole discussion ignores the problems of bridging using a measure and large extruded or large flake powders.
ps. some powders exhibit a "brazil nuts to the top" effect, ie, the large grains tend to come to the top and the smaller ones to the bottom - as such, we'll gently roll and end over end tumble the powder bottle while we're walking from the magazines to the shop [don't want that much powder and/or primers anywhere near the shop and or house]... and gently I said!