What about when the output of the lossy codec is passed through another lossy codec, e.g. MP3 through AAC over Bluetooth? I would expect better results (from the second codec) when starting from a pristine lossless source.
It's true that, technically, you'll get better results from the second codec when starting from the uncompressed source. Generally, it's always better to avoid unnecessary generation loss. That doesn't necessarily mean that you'll hear a difference since that depends on the cumulative output quality.
How many encodings does it take before a trained listener using good equipment in an ideal setting can tell?