There is a difference.
If you put all images into one sprite as an animation then all images will get loaded into the memory at the start of the layout together with the sprite object.
If you put each image into separate sprite and those sprites are not on the layout by default but are created one after another as you described (note that you need to have them on some separate layout to be able to create them with System -> Create object). Then initial memory usage will be lower as no images will be loaded.
The point is that there is no way to manually call the GC in JS, so if it happens that GC comes when you're watching 5th slide for instance, then it will release first four images from the memory as they are destroyed and nothing is pointing at this part of the memory any more. But GC may also not come during the whole slideshow. So this approach takes less memory initially and may take less in general, but it's not 100% promised... depends on GC.