We revisit the problem of phantom behaviour of effective dark energy in scalar-tensor gravity. The main focus is on the properties of the functions defining the model. We find that models with the present phantom behavior can be made consistent with all constraints, but one of these functions must have rather contrived shape, and the initial data must be strongly fine-tuned. Also, the phantom stage must have begun fairly recently, at $z\lesssim 1$. All this disfavors the effective phantom behaviour in the scalar-tensor gravity.