The problem of irreversibly harvesting from a general one-dimensional (Wiener-Poisson) jump diffusion population model is studied. For a wide class of stochastic models, the optimal strategy has a downwards local time reflection at a trigger level x*, which is typically known to be larger than in the corresponding deterministic problem if the uncertainty is Brownian. This paper shows that the presence of zero-mean jump uncertainty may or may not have the opposite effect on x*. The property of uncertainty increasing x* is related to the applicability of a comparison theorem.