something about marginalization using GTSAM

[YixFeng]

I tried to implement marginalization in my code using the same method as follows.
isam2_.update(new_factors_to_add, new_values, factor_indices_to_remove).print();

But I discovered that when the scale of optimization is larger, the time consumption of 'isam2.update' will increase a lot. I'm confused about whether this implementation of marginalization in GTSAM is correct. I guess the marginal factors are still in the optimizer not become a prior. Maybe, some steps or code I missed. Could you give me some suggestions?