In this paper, we elaborate on the proposals in the ICRU-62 report concerning planning target volume (PTV) margins for geometrical uncertainties during radiotherapy, such as variations in patient set-up and internal organ motion. According to the ICRU, these margins should be such that the planned dose in the PTV is representative of the real dose in the 'moving' clinical target volume (CTV). We demonstrate that the dosimetrical consequences of systematic and random geometrical uncertainties are fundamentally different, which should be reflected in margin calculations. The recommendation in the ICRU-62 report, to quadratically add standard deviations for systematic (Σ t o t ) and random (σ t o t ) errors to determine an overall standard deviation for margin calculations, is therefore generally not valid. Instead, a previously published recipe for PTV margin calculation, M=2Σ t o t +0.7σ t o t , does indeed account for the different impact of systematic and random errors on the dose in the CTV. If, for both random and systematic uncertainties, the internal and external errors are uncorrelated and quantified by the standard deviations σ i n t , σ e x t , Σ i n t , Σ e x t , then Σ t o t =√(Σ i n t 2 +Σ e x t 2 ) and σ t o t =√(σ i n t 2 +σ e x t 2 ). If the PTV margin thus acquired is deliberately reduced to spare normal tissues, the planned PTV dose is not representative of the CTV anymore. Therefore, we recommend to also report the minimum dose in the volume originally defined by the recipe (designated RTV, i.e. representative target volume).