Microdosimetry is the study of the stochastics of energy deposition in microscopic volumes by ionizing radiations and forms the physical basis of the quality factor, Q. The official connection was made more than 30 years ago by ICRU which defined Q as a function of the microdosimetric quantity lineal energy (y) instead of LET. Subsequently, ICRP re-defined Q(LET) to match the Q(y) of ICRU. Theoretical calculations of y-spectra are commonly based on Monte Carlo (MC) radiation transport simulations using either condensed-history (CH) or track-structure (TS) models. CH models are the most popular because of their moderate computing times but they suffer from a limited spatial resolution (um-mm). On the opposite side, TS models offer superior spatial resolution (nm-um) to the expense of high computing times. In human space missions, intermediate-energy electrons (keV-MeV) are of concern either as primary particles (e.g., in the Van Allen radiation belts) or as secondary delta rays in high-energy proton and heavy ion tracks. However, intermediate-energy electrons are only partially absorbed in micrometer-size targets, so accurate simulation of their transport is crucial to the calculation of reliable y-spectra for space radiation. The MC simulation of keV-MeV electrons is tricky due to the shortcomings of the available MC models, that is, CH models become less accurate < 1 MeV while TS models become impractical > 1 keV. Geant4 offers both CH (Standard EM, Livermore, Penelope) and TS (DNA package) physics models for electron transport. In this work we examine the influence of some user-defined CH parameters (step-size limit, SL, and tracking and production threshold energies, CUTS) to the calculation of the dose-averaged lineal energy (yD) in 1-3 um diameter spherical targets for intermediate-energy electrons. Calculations using the TS models of Geant4-DNA are also carried out and used as baseline for comparison. It is found that the influence of SL and CUT is of similar magnitude and sizeable (up to 60%). Choosing SL > diameter may lead to large errors unless very small CUTS are used. For low values of CUTS (<100 eV) and SL (<d/10), Standard EM Opt4 compares well against DNA Opt2 (differences < 20%) with a huge gain in computing time.