Thank You, for your answers!
Robot stand did not move. I would feel it, when I was checking the mastering stripes.
You're right, that's how it worked for me with other robots till now. But this case is much different and strange. Measured TCP moved by 1,5mm in X axis, but when I jog it according to Tool frame this offset gives me a huge error. Tool starts to move correctly, when I reduse the X offset almost to 0, like it was previous. It looks more like the robot starts to wrongly measure it's possition. Maybe the axies slipped a bit, but it still don't respond the answer "why it overrun the given distance". As I written in the topic, when I moved to joints's zeros by PR, the alignt looked ok. To be honest the gap between the stipes on each joint, greatly prevent from accurate examination.
Hope it's not a mechanical damage. Is it hard to examine those harmonic reducers?
Shouldn't this case give other symptoms like strange noises, shuttering during move?