Accepted answer
At 2 mg/mL a 1 mg dose is 0.5 mL — 50 units on a U-100 barrel — and no needle gauge changes that number. Gauge changes three other things: how long the draw takes, how much stays behind in the hub, and how much rubber you core out of the stopper. On the 18G scale a larger number is a finer needle, so an 18G drawing needle is coarse enough to draw quickly and coarse enough to cut a visible plug from the stopper. If you are drawing 50 units at a time, the dead space matters more than the bore: a fixed-needle barrel loses microlitres, a luer hub loses tens of them, and at 2 mg/mL each microlitre is 2 µg.
Answer first: use the largest bore you tolerate for drawing and the smallest for injecting, because the two operations have opposite requirements.
Flow through a needle scales with the fourth power of the internal radius under the Hagen–Poiseuille relation. Halving the radius reduces flow sixteen-fold at the same pressure, which is why a 31G needle draws so much more slowly than a 21G.
Stopper coring — punching a disc of rubber into the solution — is a large-bore phenomenon. An 18G or 21G needle inserted straight and fast is the classic way to do it; inserting at a slight angle with the bevel up reduces the risk.
The Hagen–Poiseuille relation gives flow proportional to the fourth power of radius, which is the quantitative basis for every gauge recommendation here.
The caveat is that no gauge choice makes a non-sterile preparation safe, and research-use compounds are not approved for human use.
Flow goes as the fourth power of radius. That is why the difference feels so large.