The performance of generalized likelihood ratio tests (GLRT) in radar detection is a well-studied subject, as is the performance of a maximum-likelihood decoder for a communication (comm.) receiver. The theoretical analysis of a cooperative radar-comm. receiver, however, remains an open problem. A radar cooperating with a spectrally coexistent communications system must decode/demodulate the coexistent comm. signal in order to conduct usual radar functions (target detection, classification, localization and tracking). This preliminary comm. signal cancellation is inherent in the GLRT where the decoding is accomplished via two nonlinear searches over the codebook; thus, rendering theoretical performance analyses highly nontrivial. It has been shown nonetheless that when the comm. signal is very strong, or very weak that radar performance is well-approximated with known results. If only moderate-to-weak, however, the comm. signal can be difficult to decode with residuals and channel estimation errors that interfere with radar operations. This paper will explore use of the saddlepoint method to predict receiver operating characteristics of the cooperative radar under such non-ideal conditions.