Cooridinates inconsistent before/after GPT binning

Hello, I used gpt binning to average three S3 imagery, and the output of binningcannot match the extent of the three input images. any comments?

the resultant imagery:

Previews of the three Input images: (a red circle shows the location of the resultant extent, using SNAP)

the coordinates of the input images:

and the coordinates of the result image:

the corresponding xml:

    <sourceProductPaths>S3_day1.dim S3_day2.dim S3_day3.dim</sourceProductPaths>

It reproduced this and it seems that the views are not synchronized correctly - the actual coordinates should be fine. Can you confirm this?

Thank you Tonio for the reply. Yes, I tested it again and seems only sync issue between views by the SNAP. the actual coordinate extent is correct.
However, the coordinates still seems problematic : with the input images as projected coordinates, for example Albers; after binning , coordinates will be Lat/Lon (?); to keep consistency with other layers, I need to reproject back to Albers using gpt reproject. But received error msg:

....10%...Exception in thread "SunTileScheduler0Standard36" java.lang.NullPointerException
        at org.esa.snap.core.gpf.internal.OperatorExecutor$OperatorTileComputationListenerStack.tileComputed(

This might be because your cache is too small. Try increasing your cache size: Tools->Options->Performance.