Protein thresholds have been shown to act as an ancient timekeeping device, such as in the time to lysis of E. coli infected with bacteriophage lambda. The time taken for protein levels to reach a particular threshold for the first time is defined as the first passage time of the protein synthesis system, which is a stochastic quantity. It had been shown previously that it was possible to obtain the mean and higher moments of the distribution of first passage times, but an analytical expression for the full distribution was not available. In this work, we derive an analytical expression for the first passage times for a long-lived protein. This expression allows us to calculate the full distribution not only for cases of no self-regulation, but also for both positive and negative self-regulation of the threshold protein. We show that the shape of the distribution matches previous experimental data on lambda-phage lysis time distributions. We study the noise in the precision of the first passage times by calculating the coefficient of variation (CV) of the distribution under various conditions. In agreement with previous results, we show that the CV of a protein that is not self-regulated is less than that of a self-regulated protein for both positive and negative regulation. We show that under conditions of positive self-regulation, the CV declines sharply and then roughly plateaus with increasing protein threshold, while under conditions of negative self-regulation, the CV declines steeply and then increases with increasing protein threshold. In the latter case therefore there is an optimal protein threshold that minimizes the noise in the first passage times. We also provide analytical expressions for the FPT distribution with non-zero degradation in Laplace space. These analytical distributions will have applications in understanding the stochastic dynamics of threshold determined processes in molecular and cell biology.