Past Zeeman analyses of spectra of cool stars have involved the comparison of observed Zeeman-broadened line profiles with synthetic profiles computed assuming a simple, one-component atmosphere, thus ignoring the several components seen in the solar atmospheres. Here we apply a physical line-transfer analysis to the more realistic situation in which the magnetic component of the stellar atmosphere is not identical to the nonmagnetic component. For the first time flux tubes are explicitly considered in detail in the determination of stellar magnetic fields. We assess the possible systematic errors incurred in the usual one-component analyses, by synthesizing profiles emitted from a two-component atmosphere and treating these as "observations" to be analyzed with a single (quiet) atmosphere. The magnetic flux tube and starspot models are based on current solar models. We also assess the possible systematic errors due to the presence of magnetic field gradients and those due to spectral mistyping (caused perhaps by strong magnetic activity). We conclude that substantial effects are possible which render the results of one-component analyses somewhat uncertain, with systematic errors potentially as great as 40% in the derived magnetic flux. However, it appears that two-component models consisting of quiet and flux tube regions may be oversimplified, since the predicted line profiles tend to be shallower than those observed.