We generalize and extend the recently proposed method [1] to account for contributions of system size (or volume/participant) fluctuations to the experimentally measured moments of particle multiplicity distributions. We find that in the general case there are additional biases which are not directly accessible to experiment. These biases are, however, parametrically suppressed if the multiplicity of the particles of interest is small compared to the total charged-particle multiplicity, e.g., in the case of proton number fluctuations at top RHIC and LHC energies. They are also small if the multiplicity distribution of charged particles per wounded nucleon is close to the Poissonian limit, which is the case at low energy nuclear collisions, e.g., at GSI/SIS18. We further find that mixed events are not necessarily needed to extract the correction for volume fluctuations. We provide the formulas to correct pure and mixed cumulants of particle multiplicity distributions up to any order together with their associated biases.