Just to be clear, the 'old' raster(s) is/are NOT in a Mosaic dataset currently, and the old raster is in one large (seamless) raster. The 'new' rasters are tile-based and have already been loaded into a Mosaic dataset.
Here is my recommended solution:
- Create an 'old' Mosaic dataset from the 'old' seamless raster using the tile boundaries from the new Mosaic so that the tile boundaries match.
- Once you have two distinct Mosaic datasets (one 'old' and one 'new') then you can use the Synchronizing mechanism as described above.
- Lastly, you can replace the original seamless imagery layer with a layer pointed at the updated 'old' Mosaic after the tiles have been synchronized.
An alternative solution would be to use Raster Math to:
- Merge the tile-based updates into a single raster
- Create a 'mask' raster where new tiles are present in the merged result
- Null out the mask areas from the original single imagery layer
- Add the raster with the updates and the masked original imagery layer
The results will be a single imagery layer with the updates "knocked into" tile areas where they are present. Of course, then you could create a new Mosaic and load the imagery into it so that you don't have this problem in the future.
Lastly, keep in mind that by loading the imagery into a mosaic you are simply trading one set of difficulties for another -- is it best to manage a large single ECW or a mosaic where tile boundaries could shift over time?