In optical communication systems based on photon-counting protocols, using highly efficient superconducting nanowire single-photon detectors (SNSPDs) is currently considered an effective solution for achieving high communication rates. However, in the process of increasing the communication channel throughput by reducing the slot width, as the slot frequency increases to the GHz level, the sub-nanosecond-level jitter caused by detection systems becomes non-negligible relative to the entire signal slot and greatly affects the accuracy of slot synchronization. To accurately characterize the photon arrival distribution of SNSPDs and further reduce communication error rates, we constructed a system. Through data analysis, we discovered that the photon arrival intensity distribution under the influence of jitter conforms to a non-homogeneous Poisson process (NHPP) and proposed a correction model for the photon arrival distribution and provided the corrected expression for the maximum likelihood estimation (MLE) algorithm. In subsequent experiments, combined with a blind search strategy, we enhanced the robustness of the model under different slot offset states and reduced the hardware complexity of existing synchronization schemes. The results show that at signal photons Ks=1.624 and background noise photons Kb=0.016 per symbol, the ideal Poisson model's hard decision error rate is 10.56 %, while the corrected model achieves an error rate close to the ideal synchronization performance at 8.9 %.