I believe it is losses inflicted during battle / losses suffered during battle. So it is combat efficiency, the kill/loss ratio. It does not include losses due surrender/encirclement. It does also not include losses due attrition.
What is combat efficiency? A sledge hammer to squash a bug? Firstly, there is no universal agreed definition and even different parts of any military will judge it differently. But if I and my nine friends beat up one single opponent (we'll assume all the losses were on his side) was our side efficient?
Efficiency is not force. Efficiency is result achieved by minimum effort exerted. Hence, any graph like that (MP losses inflicted/MP losses taken) does NOT measure efficiency but is only a further cute representation of the info already presented in the other two graphs.
An example of the various concerns of judging battle efficiency was the Russian way of counting how many horses they killed (for food I assume) for each German tank destroyed.
Further, a true efficiency would most definitely put the greatest focus on enemy MP losses suffered DUE TO SURRENDER/ENCIRCLEMENT because there is nothing more efficient than that - SURRENDER. The fact that the graph omits the most important factor of battle efficiency and skews total enemy MP lost makes the graph probably not even worthy for a storyline commander to display!
And because attrition is such a major factor of the game, it should be shown (but not in that graph). While attrition suffered during a war most certainly is a measure of battle efficiency in every way (planning, logistics, adaptation, flexibility and lack of control) I quite agree with Mr_BOnarpte that it should stay out of a "battle efficiency" graph only to simplify creating a better graph than currently shown AND BECAUSE seeing the attrition figures separately probably is more interesting (and useful) to players - especially as a chart connecting monthly points.
Saying "kill ratio = combat efficiency" is a statement expressing failure to understand how wars are won. The statement is much more applicable to Cro-Magnon times when each caveman possibly had the same length club. To be precise, "kill ratio = exactly kill ratio", and nothing else. But "battle efficiency" is an idea that is totally concerned with the idea of winning, weapons working better, and does not need necessarily to include killing (although that is often the side effect of sinking a ship with more efficient torpedoes, etc). It could include simply rendering the enemy ineffective, such as arranging their surrender or denying them oil for their equipment.
If USA with an army of say 1 million achieves a 1:10 kill ration ratio against China's army (let's say 100 million) we can easily see it will be the most inefficient war the USA might ever fight. In fact the USA would be out of soldiers when China has 90% army left. A nuke would be much more efficient, as would a USA retreat. And if the USA did nuke China, would anybody give any credence to a claim that the USA just achieved a kill ration of zero-to-a-million? No, because kill ratio happens on the battlefield individually between soldiers while things like artillery tend to skew the true facts.
But battle efficiency begins in the minds of strategists, moves thru armaments factories, and eventually tests itself in situations with unlimited factors to decide a winner if the game is played to a conclusion. Time that a war takes would be a much better measure of battle efficiency than any body counts. So, it can be correctly stated that "A 7-day war to create Vichy France must be very battle efficient".
But as regards AoD, player working the situation to get surrounds and over-runs is probably the most important measure of battle efficiency; and it is quite wrong that the graph excludes that but calls itself what it does.
