GOF testing and time intervals

[cassowary43]

I am running a CJS model with 7 capture events that differ in the number of days between these capture events (13, 18, 9, 32, 17, and 21 days). I have set the time intervals three different ways:
1) entering the actual # of days (which I presume gives me a daily survival estimate)
2) # days/14 days (entered as a decimal; gives a 2-week survival estimate).
3) # days/28 days (entered as a decimal; gives a 4-week survival estimate).

I am generally happy with the results...all provide the same dev, df, AIC, etc in the global model. However, when I test this global model for its lack-of-fit using bootstrap GOF, I get very different results in the bootstrapped estimates. The trend is this: as I divide the sampling interval into a larger time frame (i.e. 1 day vs 14 days vs 28 days), my global model shows worse and worse fit.

Is there an explanation to this? What would people recommend doing with my time intervals? How much should I even worry about this GOF testing? As I adjust c-hat between 0.5 and 2.0 (I haven't gone further than that), model rankings do not change. Also, I would like to point out that daily estimates in this case are virtually meaningless, as they are around 0.99 and differ minimally among treatments. If daily estimates are the way to go, is it a problem to report these estimates coverted to 28 day estimates?

Thank you very much!
Erik

[cassowary43]

Does anyone out there have any ideas? Is anyone even considering my questions? Am I just way off??

Thanks,
Erik