This paper is a new step towards getting rid of nonlocality from quantum physics. This is an attempt to structure the nonlocality mess. "Quantum nonlocality" is Janus faced. One its face is projection (Einstein-Lüders) nonlocality and another Bell nonlocality. The first one is genuine quantum nonlocality, the second one is subquantum nonlocality. Recently it was shown that Bell "nonlocality" is a simple consequence of the complementarity principle. We now show that projection nonlocality has no connection with physical space. Projection state update is generalization of the well known operation of probability update used in classical inference. We elevate the role of interpretations of a quantum state. By using the individual (physical) interpretation, one can really get the illusion of a spooky action at a distance resulting from Lüders' state update. The statistical interpretation combined with treating the quantum formalism as machinery for update of probability is known as the Växjö interpretation. Here one follows the standard scheme of probability update adjusted to the quantum calculus of probability. The latter is based on operating with states represented by vectors (or density operators). We present in parallel classical and quantum probability updates. From this presentation, it is clear that both classical and quantum "faster-than-light change of statistical correlation" take place in mental and not physical space.