Thanks for your help. But I’m still a little confused how Ellipsoid-correction > Geolocation-Grid should work. SNAP gives location of each pixel before geometric correction. Coordinates are in Tie-point grids. I checked the position of a few characteristic pixels along shoreline before and after geometric correction. Before correction location is much better than after (fits Sentinel 2 data and other maps of this region).
If the geometric displacement is caused by the incidence angle and the topographic variations how Tie-point grids were calculated? Are they used by Ellipsoid-correction > Geolocation-Grid or this function calculate coordinates independently?