Hi,
I did the tests:
* Replacing the decoder with the dummy and testing in DC the behaviour is exactly the same: At full power starts to run very slow and stops after few centimetres.
* Remove the motor and feed directly: Motor works perfect
So if it's not the motor neither the decoder, the problem is in between. And in between there is the worst design and concept I've ever seen in a loco. Now I love the KISS (keep it simple and stupid) philosophy of Kato!!!!
I try to explain: Decoder is plugged in a small circuit that also has a reel and some more elements. This circuit has a wire to feed the positive from the wheels, and it has no black wire to feed negative

Finally I saw negative current is taken from the screw used to fix this circuit. This screw is in touch with a metallic part of the chassis, and that last is in contact with the other side of wheels. And even worst, the motor is soldered with a wire to this circuit (negative), and positive is taken by contact with this circuit.
So there are so many things that can fail in the design, that I will remove that circuit and solder a wired decoder directly to the motor and wheel feeders, as I do with Kato locos.