Modeling t models in Pradel for short (6-years) data sets

[Robertal]

When modeling lambda in the Pradel model, is it valid to consider a little-t model for short (5-7) year data sets? My question arises, because the the first 2 and last esitmates of lambda are not valid, and so (with a 5-year dataset) we are really only obtaining one estimate of Lambda. Does the little-t model get artificially selected because the lambda values for these non-valid years are all over the place? If so, at what point (or length of data set) is it appropriate to consider a little-t model for lambda with Pradel?