(I have edited my question in case someone can benefit from my experience)
The data
Capture-recaptures for male and female rodents (Oryzomys palustris). 13 primary periods with 3 secondary periods each.
The problem
In summary, interaction models (e.g., S(Sex*t)) resulted in male survival rates near 1.0 for every period. In addition, female emigration rates in most models were not inestimable.
What the problem appears to have been
First, the data for this particular period was more sparse than data from other periods and it was simply not possible to estimate the additional variation in interaction models. Secondly, there was only one single occasion when a female was unobserved between primary periods and therefore there was no temporary female emigration.
My 'solution'
I did not use interaction models and I fixed female emigration rates to 0. Parameter estimates were decent and model selection turned out to be insightful. My lazy mistake was that I made assumptions about this small analysis without having first looked at the M array to get a better feeling for the data. Lesson learned.
Cheers,
Miguel