What if the data are not Poisson distributed?

That is not a problem. Deviations from Poisson distribution can be taken into account by checking overdispersion in TRIM. Incorporate overdispersion in the estimation procedure, else the standard errors might be seriously underestimated.