Average area calculation differences between raster stack and single Maxent models

I ran 100 bootstrap replicated models, for which I want to calculate the area of suitable habitat based on three different threshold rules (MTP, ESS and MaXSS).

The issue I’m having is that for the same threshold I get different mean area values based on either the entire raster stack of 100 models than on area single models summed and then averaged.

For example:

When I create an average raster stack of all model replicates (100 models) and reclassify it using the average threshold value of MTP from the maxentResults file I get 11 000km2.
However, when using an iterative process where I iterate through each replicate and reclassify it with its corresponding threshold value from the maxentResults file and then calculate the mean on the sum of the area from all replicates I get something totally different in this case 17 000km2.

Could someone explain why I get different mean area values?

What I would like to achieve is to create boxplots showing the variation in the area of suitable habitat produced from each model replicate.