Ab initio result: particles count

Dear all,

I am a new user and I’ve noticed that after an Ab Initio Reconstruction run (even with 1 class) the number of particles is lower than in the input of the same run. If I have understood correctly it applies a threshold removing all particles with a probability lower than a specific value. Unfortunately my dataset is quite heterogeneous and with this set up I have lost many particles belonging to the complex of interest and I would like to avoid it.
I thought it was the class_probability_threshold parameter because it is set to 0.9 so I tried to change it with 0.5 and 0.1 but nothing, I got always the same result.
Can you help me with that?
Thank you