Dealing with classes with defects (flexibility)

Hi All,

I am refining a dataset of a multi-subunit complex ( A + B + …). After heterogeneous refinement, I got two good classes. Although in one of the two classes, A is good, but B has some parts relatively weaker, while in the other class, B is fine, but A has weak density. For this case, is there some way to combine the good parts of both A and B together to generate one map with good density for both A and B? Alternatively, if I model and refine A and B against two maps from the two classes, respectively, how should I present it in a publication? I also tried combining both classes, and this resulted in averaging everything, with relatively weaker A and B than other parts. One way may be to do masked refinement (or local refinement in cryosparc) with the combined classes, but I am wondering if there is other strategies I could try? Any comment is appreciated.