I arrive at a similar outcome, though by looking at world balance sheets for the respective commodities.
Ratios can be really useful for timing entries/exits, realising that when exceeded they can generate swapping between commodities by end users, which is the mean regression at play that you mention.
Same things happens in Australia every year with the ASW/F1 ratio and the associated swapping between feed rations that feedlots will use, depending on that spread. e.g. long F1 against ASW1 was an obvious move in December when the spread was $65 and there ended up being $30+ in that play.
