registration got unmatched images

Hi, here I tried to register the DAPI staining images between nanostring and xenium, and apply that transformation to the segmentation masks, the images aligns well in the napari, however, when I extracted the registered images and plot with matplotlib, I found they were not matched, do you have any suggestions? Thanks!

```python
      xenium_sdata = SpatialData(images=img_xenium,shapes=shape_xenium)
      nano_sdata = SpatialData(images=img_nano,shapes=shape_nano)
      affine = align_elements_using_landmarks(
          references_coords=xenium_sdata["landmarks"],
          moving_coords=nano_sdata["landmarks"],
          reference_element=xenium_sdata["DAPI"],
          moving_element=nano_sdata["DAPI"],
          reference_coordinate_system="global",
          moving_coordinate_system="global",
          new_coordinate_system="aligned",
      )
      affine
      def postpone_transformation(
          sdata: SpatialData,
          transformation: BaseTransformation,
          source_coordinate_system: str,
          target_coordinate_system: str,
      ):
          for element_type, element_name, element in sdata._gen_elements():
              old_transformations = get_transformation(element, get_all=True)
              if source_coordinate_system in old_transformations:
                  old_transformation = old_transformations[source_coordinate_system]
                  sequence = Sequence([old_transformation, transformation])
                  set_transformation(element, sequence, target_coordinate_system)
      
      postpone_transformation(
          sdata=nano_sdata,
          transformation=affine,
          source_coordinate_system="global",
          target_coordinate_system="aligned",
      )
      img_nano_aligned = nano_sdata.transform_element_to_coordinate_system('DAPI','aligned',True)
      img_xenium_aligned = xenium_sdata.images['DAPI']
```

Here are the images that I got with matplotlib
Screenshots
img_nano_aligned

img_xenium_aligned

aligned images in napri(nanostring DAPI is colored with red and xenium DAPI is colored with blue)

Start by tracing align_elements_using_landmarks and transform_element_to_coordinate_system for the DAPI image path, then compare the coordinate systems and transformations used by the napari and matplotlib outputs. The issue is done when the extracted registered images match the alignment shown in napari, with a reproducible example or a documented explanation if the behavior is expected.