Chao's models in CAPTURE

[cebert]

Dear all,
for purpose of comparison I tried some CAPTURE population estimates in addition to MARK closed captures (and full closed captures with heterogeneity). I used two unfortunately quite small data sets with low recapture rates. Now the 'best' models in both MARK and CAPTURE always assume time dependence and individual heterogeneity, which of course is not really a surprise. But what puzzles me when comparing the results of the different models is the fact that it is always the Chao models (Mh and M th) in CAPTURE which yield unexpectedly large estimated N compared to all other models (either Mh Jackknife in CAPTURE or the MARK Mh and M th models). Now as far as I know, the Chao models are said to be especially suitable for small, sparse data sets, but their results with our data seem unreasonably large.
Has anyone of you an idea or a hint - or some own experience - how to interpret these results? I would like to have a hint how this large difference is explained and which models now are suited for our data (all other MARK and CAPTURE models yield rather similar results - it's only the Chao models which differ so much)...
Greetings and thanks in advance for help!
cebert