Wow. Uncheck a box. So simple I missed it. Will give that a try soon. Not as ideal as a raw AAC stream – I’ll need to get creative with my muxing scripts – but it is certainly workable.
Does anyone know the answer to the more detailed part of the original question?
Namely: In order to support iPad and other iOS devices, must the AAC stream be the first audio track in the container AND must it be stereo only?