For insertion-type lithium-ion batteries, the solid-state diffusion coefficient of Li$^+$ in an active material is considered a key parameter within the research community. The capabilities and limits of related parameter extraction methods are usually well-established. However, there is a gap in understanding the influence of the applied measurement setup. In practice, many setups unintentionally violate the assumptions of the extraction method. We apply the galvanostatic intermittent titration technique (GITT) in virtual experiments using 3D microstructure-resolved simulations in varying model measurement setups. Diffusion coefficients are extracted by applying a state-of-the-art Bayesian optimization approach which is particularly suitable for non-uniquely solvable problems. The investigated parameters are within the typical literature range of LiNi$_{0.8}$Mn$_{0.1}$Co$_{0.1}$O$_2$ (NMC811). Because experimental boundary conditions are precisely known within the simulation setups, the influence of microstructural features on the extracted diffusion can be isolated and quantified. The investigation shows, that the applied monodisperse thin-electrode and single-particle setup are capable of extracting the actual diffusivity with up to 4% and 17% point-estimate deviation, respectively. Particle cracking proved to have the largest impact on extracted diffusion coefficients. Nevertheless, all predictions remained close to the correct order of magnitude, i.e. point estimates deviated at most by factors in the range of 10$^{±0.88}$.