CQRS solves one problem for sure - it separates domain models from read models. So with properly implemented CQRS you would not run into this kind of problems and your domain models would be focused about business logic, not about reading outcomes of it.
how would you reconstruct an Aggregate Root from the Read-Model?
You don't :) You don't; you just use id's from your read side to get domain model with same id from your command side.
