While developers and users of modern software systems usually only need to interact with a specific part of the system at a time, they are hindered by the ever-increasing complexity of the entire system. Views are projections of underlying models and can be employed to abstract from that complexity. When a view is modified, the changes must be propagated back into the underlying model without overriding simultaneous modifications. Hence, the view needs to provide a fine-grained sequence of changes to update the model minimally invasive. Such fine-grained changes are often unavailable for views that integrate with existing workflows and tools. To this end, model differencing approaches can be leveraged to compare two states of a view and derive an estimated change sequence. However, these model differencing approaches are not intended to operate with views, as their correctness is judged solely by comparing the input models. For views, the changes are derived from the view states, but the correctness depends on the underlying model.This work introduces a refined notion of correctness for change sequences in the context of model-view consistency. Furthermore, we evaluate state-of-the-art model differencing regarding model-view consistency. Our results show that model differencing largely performs very well. However, incorrect change sequences were derived for two common refactoring operation types, leading to an incorrect model state. These types can be easily reproduced and are likely to occur in practice. By considering our change sequence properties in the view type design, incorrect change sequences can be detected and semi-automatically repaired to prevent such incorrect model states.