I know and understand all this; clearly we are talking at cross purposes.He does say (including in one of the bits I quoted) that some commercial mills test the ingredients.
For people who cannot test the ingredients, the obvious workaround is to just work with the worst numbers for each ingredient. Use the lowest number for things like protein, and the highest number for anything that is dangerous in excess. If both ends of the range matter (like for salt), do the math both ways and make sure it is within the safe range. Feed manufacturers could do that too, although it is probably cheaper for them to test the ingredients than to maybe use extra of the more expensive ones, given that they are making large batches of feed.
In the USA, animal feeds have a "guaranteed analysis" tag that lists the minimum of some things (example: protein) and the maximum of some other things (example: fiber.) Some things list both a maximum and a minimum (example: salt). I assume the manufacturer is free to exceed those numbers, but can have trouble if they are caught short--at a bare minimum, it would violate laws about false advertising, to "guarantee" something and then not actually provide it.
The point I keep trying (and evidently failing) to make is that current research (from the FAO, as well as people like Spector) is emphasising that we have to start paying attention to the QUALITY of the nutrition in foods, not just the quantity. If, say, 40% of the protein in a given ingredient is not bioavailable (digestible, metabolizable) then in the consumer's gut it is actually a great deal less nutritious than it appears to be in the lab. Simply focusing on the numbers on the bag, as if what comes out of the assay is exactly what is absorbed into a body, is not enough. That's what allows things like melamine and dioxin to pass these assays while poisoning consumers.
And then it turns out (surprise!) that everybody's digestion is different. So what number comes out the assay or is found to be digested in a given case is not necessarily representative of what happens in any other body. The fixation on number crunching ignores this.