Well, imagine shooting 10 cm/px GSD, and your GCP center to 10 cm radius differs by 3-4 cm. You have 1 px for this, so there must be some kind of averaging happening.
I guess we could generalize from the, and say the smaller the GSD value, the more critical it is to have a even/flat target. So, if doing volume measurements, this would be a good idea! If doing i.e. a glacierarea of multiple square km at 10 cm/px GSD, then it’s doesn’t matter much.