Why is the number of population for GA's first single run random?

Before I switched over to GA to utilize the mixed integer operation, I was using gamultiobj.As I save the input parameters in my custom fitness function, I know that gamultiobj would first run 1 population of the first generation and then run the rest of the population in first generation, before continuing to run all populations of future generations at the same time.

By tracing through ga.m, I see that GA goes through a similar process. However, in my current setup, I set the run to contain 150 population and I noticed the first run would have a random number of population. First time it had 10 population, then 15, then 15 again. The magic seems to be happening in gaminlp, but I don't have access to this function.

I know that GA with integer constraint should not have Aeq and Beq, but I do have A and B, which shouldn't have any issues as it is not mentioned in the documentation.

Does anyone know why GA does this?

Thanks,

Jesse