We introduce a new method to measure structure factors from parallel beam electron diffraction (PBED) patterns. Bloch wave refinement routines were developed which can minimise the difference between simulated and experimental Bragg intensities via variation of structure factors, Debye parameters, specimen thickness and -orientation. Due to plane wave illumination, the PBED refinement is highly efficient not only in computational respect, but also concerning the experimental effort since energy filtering is shown to have no significant effect on the refinement results. The PBED method was applied to simulated GaAs diffraction patterns to derive systematic errors and rules for the identification of plausible refinement results. The evaluation of experimental GaAs PBED patterns yields a 200 X-ray structure factor of -6.330.14. Additionally, we obtained -6.350.13 from two-dimensional convergent beam electron diffraction refinements. Both results confirm density functional theory calculations published by Rosenauer et al. and indicate the inaccuracy of isolated atom scattering data, which is crucial e.g. for the composition evaluation by lattice fringe analysis.