An important element of most modern spectrographs is their ability to take spectra of many objects simultaneously using either multiple slits or optical fibers. While this is an extremely efficient method of utilizing telescope time, it comes with a price of the much higher level of preparation in advance of your run, and time for setup at the telescope (both during the day and while observing).
In order to prepare for observing multiple objects, you need to have an object list with high-quality relative astrometry (better than 0.3'' is usually required). Most observers do not bother to observe astrometric fields, and hence their data are often not of a sufficient quality to do the kind of astrometry required for multiobject spectroscopy.
In order to illustrate this point, imagine that you wish to observe faint galaxies along a strip with COSMIC in order to obtain redshifts. You choose to use a 1.5 arcsec slitwidth, which is typical for this application. You want to keep the objects on their slits to within 0.25 arcsec rms, because a 3 deviation would then put a given object halfway off the slit. This means that you need to know the plate scale and distortions in your preparatory CCD imaging to an accuracy of 0.25 / 240 = 0.1%, to cover the full 4' = 240'' perpendicular to the slits. So if the pixel scale on your CCD is arcsec, then you would need to measure that scale to an accuracy of 0.0003 arcsec. If there were additional distortions in your field-of-view (extremely likely!), then you would have to characterize these fully across the full field to this same accuracy. Needless to say, many observers do not bother themselves with measuring these effects to the required accuracy.
An excellent way to characterize the plate scale and distortions in your input images is to observe one or more star field which has many stars with high-quality coordinates (like well-studied globular clusters).