Are there any guidelines as to how accurate cable lengths should be when conducting arc flash study? It seems the be that cables usually have a relatively low impedance when compared with transformers, especially when dealing larger cables. When plants don't have decent drawings, collecting cable lenghts can be quite time consuming. Is eyeballing it to get a relative length sufficient?

Estimating length is as good as is possible. Choose a method that works for you whether that be be pacing-off or measuring the length on the surface for underground runs, or measuring from plan drawings; with added footage for the risers into equipment.

Consider that adding cable length will at some point cause risk category to go UP if the added impedance causes instantaneous trip setting to not pick up. That particular footage of cable length can have a dramatic effect on arc flash calculated exposure.