One thing, and actually the most important thing about FE that hasn't been mentioned too much, and not by some folks at all, is that what breed is most efficient is nearly entirely due to the period of time we measure them for, or the market endpoint at which we measure them for.
In other words, high gainers are generally most efficient over a test period that is measured in days, fat cattle are most efficient to a fat endpoint, and lean cattle are most efficient to a retail product endpoint.
Anytime we expect a biological type to perform to a "different" specification than they are designed for, we lose efficiency.
This is why crossbreds tend to excel-they are matched to both marbling and retail product endpoints, and they can bring gain in form both sides of the equation, plus they get a small boost from heterosis.
So, the argument for GAIN indicating FE covers less than 25% of the argument, and GAIN can only be used to improve FE up to the point that mature cows weights increase beyond environmental constraints.
In other words, for the REAL ranchers around the USA, they have used up about all the GAIN benefit they can get. Farmers can haul more grain and feed to the cows, but ranchers don't or can't afford to.
Anyone still chasing gain to achieve efficiency is completely lost, unless of course their cow MW EPD are below 0 and their YW EPD are below 30-40. In that case, they might still want to select for more gain to improve efficiency.
For ranchers who have their production maxed out with low MW and 40 pounds of YW, someone will hopefully come to their aid in the future in terms of delivering seedstock that are efficient, even if they are smaller.
Variation in FE occurs at all rates of gain and at all mature sizes. To use ADG just eliminates some of the variation, but not all of it, and it only captures the variation that is useful for more high-input environments.
Badlands