Many revisited mortality trees don't have DIA recorded (at least in western states), which leads tpa() to do some dramatic under-counting when type="all" or "dead". I can understand the reasoning behind this choice, especially when estimating anything other than TPA. One solution would be to include these trees in TPA estimates, and use PREVDIA (with a disclaimer) when necessary.

Possibly not a problem many people will run into, but thought it worth pointing out!

Thanks,
Daniel Perret