Hello,
The components are created during the alignment process. So if images do not have sufficient overlap, or the common features are not recognizable, it could be that multiple components get created from the same alignment process.
About the models. Every time you alter in any way an existing model, a new model will be created. If you filter out some parts, smooth it out, simplify it, or do anything to change the triangle count and the topology, a new model will be created.
So if you classified your model, these classes won’t be projected onto a new model. The reason is that the classification is connected to the topology of a model, not to the component. The process behind the classification is that it selects certain triangles of a model and places them under a classification layer.
I will link this thread here too, since it is the same question.