The milk epd is complicated. Some epd's (bw, ww, yw) are based on easily measured traits (pounds of weight) from the animal and their progeny. Expectation is that more pounds measured will result in higher epd's for the trait.
Milk is not directly measured. I have simmental. Their units of milk epd is pounds. Pounds of weaning weight due to the mother's milk. Genetic growth ability will affect weaning weight. Mother's milk (quantity and quality) will also affect the calf's weaning weight. This implies that the weaning weights submitted on progeny will affect the ww epd of the bull as well as the milk epd of the bull. In the case of milk, it would have to be the weaning weight data of calves from daughters of the bull. The question I have always had is how does the epd model split the contribution between growth due to genetics (ww epd) and growth due to mother's milk (milk epd). If a calf weans off good, is it due to growth ability or mother's milk ability? Well, both, but how much to attribute to each? I don't know the answer.
I remember a 2 y/o heifer than weaned a heavy bull calf. The calf's milk epd came back significantly less than the average of the sire and dam's milk epd. My thinking was that the dam must have been a good milker to wean a heavy weight calf. Response from the association was that growth and milk are antagonistic. That high growth animals tend to have lower milk ability. Since the calf's growth was more than expected, the milk epd was reduced. I was and still am confused on that.
Simmental association registers crossbred (percentage) animals and calculates epds for them. So, heterosis comes into play in the calculations. A Sim-angus cross should have increased weaning weight due to heterosis. So, a small portion of the actual weaning weight on the calf is discounted since it came from the heterosis, not the genetic potential of the sire or dam. Complex and confusing.
Enhanced epds from the genomics is supposed to increase the accuracy of the epd's equal to the increase of around 25 progeny. Based on my limited understanding, I tend to think that the new epd might be more correct than the sire/dam average.
It seems to me that epd's on young animals are much more likely to go down than up as they get older and epd's get more accurate. Just seems like half would go up and half would go down if the science were correct and accurate. Complex and confusing.