They are not "fake accessors", they are object-oriented accessors, as opposed to the usual method-oriented accessors (neither of which couples the interface to the implementation). Properties of an object are a legitimate part of the object model, not merely part of its implementation. They are conceptual objects. Method-oriented design (where the object name is embedded in the name of a parent object's method, or is passed to it) sweeps that under the rug. There is reason to do that; I'm not saying it shouldn't be done. I just think people should be aware of why it's done. Because it rankles.

Accessors certainly can indicate poor design, but they don't always. Some objects are intended to be freely manipulated. The objects in a drawing program, for example. Or the accounting books at Enron. The object has to model the process, it can't dictate it.