In order to visualize the functionality of a biological cell, it is often desirable to label specific proteins. In this work we concentrate on the optical theory of visualizing colloidal gold labels with soft X-ray microscopes, where scattering from small gold spheres used as labels dominates the image. Using numerical simulations of bright and dark field imaging, we compare different configurations of condenser and objective lenses in transmission X-ray microscopes, and configurations of detector and objective lens in scanning transmission X-ray microscopes. It is verified that the contrast of small, closely spaced features is strongly affected by changes in these configurations; the optimum situation is to have the condenser aperture (in TXM) or detector aperture (in STXM) equal to 3/2 that of the objective numerical aperture.