Hi, mvkolos!
You need to split RGBD data into two separate sets of images: RGB images and Depth images. Then provide these two separate sets to the Metashape.app.document.chunk.importLaserScans python command. Depth images paths should be passed to the "filenames" argument, and RGB images paths should be passed to the "color_filenames" argument. image_path argument should point to some existing directory and end with "{filename}.tif". Metashape will substitute it with an actual file name.
Depth images are preferred in floating point .tiff or in .png formats. If it doesn't work - please attach depth and color images, so we could check them.
Misleading "Embedded depth data is not found for the photo <my_depth.tif>" is appears as we try to interpret the data as images from smartphones, but it is not the case.
Note that depth and color images should have different paths.