Yeah, trying to infer probabilities from a dataset with one sample point is strange. (Especially when all our priors are from philosophical axioms and not observation.)