Exactly which method do you mean? The original SCS-CN method was designed for an event quick runoff estimation, therefore it is not suitable for any groundwater-related calculations.

NRCS-CN is based in runoff coefficients that are dependent upon land use and drainage class. So any error in evaluating each could extend to CN value. In addition soil moisture defines the runoff coefficient . This also can introduce uncertainty. The CN itself introduces error if textbook values are used. There s large variability in these values. In ungauged watershed it is difficult to calibrate, thus these errors can be severe. 

Ground water balance means you need data on soil types and infiltration rates. At the first step, CN requires land cover classification where many errors may occur during the classification process. This, however, depends on you study area homogeneity and complexity.