In this publication, uncertainties in and differences between the MC@NLO and POWHEG methods for matching next-to-leading order QCD calculations with parton showers are discussed. Implementations of both algorithms within the event generator SHERPA and based on Catani-Seymour subtraction are employed to assess the impact on a representative selection of observables. In the case of MC@NLO a substantial simplification is achieved by using dipole subtraction terms to generate the first emission. A phase space restriction is employed, which allows to vary in a transparent way the amount of nonsingular radiative corrections that are exponentiated. Effects on various observables are investigated, using the production of a Higgs boson in gluon fusion, with or without an associated jet, as a benchmark process. The case of H+jet production is presented for the first time in an NLO+PS matched simulation. Uncertainties due to scale choices and non-perturbative effects are explored in the production of W ± and Z bosons in association with a jet. Corresponding results are compared to data from the Tevatron and LHC experiments.