The problem of testing a point null hypothesis in Poisson distribution is considered. The Bayesian measure of evidence against the null hypothesis is compared with P-value (or observed significance level). Lindley and Jeffreys approach is followed for Poisson distribution. Bayesian measure of evidence against the null hypothesis expressed in terms of the infimum of the posterior probability of the null hypothesis when the infimum is taken over the class of some reasonable priors. The result shows that the P-value is much lower than the lowest probability obtained even if all priors are considered. According to the authors this irreconcilability between P-value and any Bayesian analysis of null hypothesis is no longer valid for one sided null hypothesis.