Is it better to use few presence location without bias ? Or many ones but with possible pseudo-replicas ?

[Ulysse Faure]

Dear Group,

I would like to run a Habitat Suitability Model (HSM) in Maxent for brown bear den-site in a 3400km2 mountain area in Romania, but my presence location dataset is unfortunately not the best.

I have ca. 120 dens location in my study area, that were monitored in different ways : 9 were found thanks to collared bears and the rest was found thanks to forestry services, hunters and villagers. 

My question is, I don't know if I should only use the 9 dens locations found thanks to collared bears (each den come from different bears, so den locations are independent), or if I should consider the entire dataset (with possible pseudo-replicas, with more than one den per bear, so a possible bias in the results and with most of the dens monitored in a small area) ? 

I know that Maxent can deals with very small sample size, but I never used this algorithm before, so if you have any advices I will be very grateful.

Thanks 

Ulysse