Estimates of pi when using Mixture models in MARK

[devcharan]

I have a quick question regarding the model {p(.)=c(.), pi} in
MARK...with many datasets, the pi estimate is exceedingly small (e.g.
0.57 E -17), so for all practical purposes the estimates are identical
to model {p(.)=c(.), pi=1}, with one redundant parameter being
estimated.

My question is, why does this happen? Am I doing something wrong, or
are the mixture models unstable in any way?

This has happened repeatedly, even when the data are not too sparse (10-15 occasions; approx. 10-30 individuals).

Thanks in advance for your help!