There are many ways to select cattle. Phenotype, cow family, pedigree, breeder reputation, epd's, ratios, etc. But there are perhaps some pitfalls in all of those.
In the example above, the attitude was "the more milk the better". I think that is where the trouble started. Let's think of a scenario where someone wants more growth in their herd. They go to a seedstock producer to buy a bull and tell the owner "I want to buy the highest weaning weight bull you raised this year". Sort of like "the more milk the better". Guy pays for that bull and turns him out with his cows. He selects his replacement heifers based on the same criteria - highest weaning weights. Since he wants more growth in the herd. Next bull is purchased based on the same criteria. Repeat that for 10 years. What are the results? Probably high growth cattle. But what about the fertility? Udders? Longevity? Structure? Disposition? Profitability? You get the idea. The thought that "the more milk the better" was perhaps more of a poor decision than using an epd. Long term single trait selection will cause a single trait to excel at the detriment of a balanced efficient cow herd.
Epd's are a good tool, but selecting for extreme epd's as your main criteria is perhaps similar to the charolais story above. Maybe the risk is more so in application of a tool than the tool itself. Whether selecting for phenotype or epd. May be easier to have a more moderate balanced trait herd than a herd that is tops in every trait. May be more profitable as well.
Yearling bulls are always going to be somewhat of a risk. They are unproven. A consistent cow herd (breed composition, size, growth, milk, etc) may make it easier to select a bull that works. When there is a bunch of variation in the cow herd, that probably makes selection of a bull more difficult.
Then epd accuracy is low for most animals. Given that, a bull in the top 30 percentile may not be superior to one in the top 70 percentile. But epd's in the bottom 10% for multiple generations back may be a flag.
I think I understand and agree with what you said. There are in my opinion no totally foolproof methods or philosophies about breeding cattle in every circumstance.
I remember the days of selecting calving ease bulls based on smooth shoulders, under 100lbs birthweight, unassisted from a heifer etc.
EPD's helped to weed out some guesswork and could be backed up fairly quick for CE.
The example I gave of Charolais was coming in at a time when EPD's were just getting going and Charolais were behind the curve on utilizing them until an opportunity arose. The breed at that time was heavily influenced by the show ring movers and shakers. The champion bloodline cattle were streamlined to just a few lines and heavily sought after. The daughters from those lines didn't milk near enough if at all.
Suddenly milk was important again and people were clamoring for bulls with good milk EPD's. Those numbers we're very fluid, there may be more consistency now.
I've also found that pedigree history is a factor for traits too, but when mating animals with traits at opposite ends of the spectrum I call them fire and ice matings, a lot of variation happens.
An example the Angus bull Bismarck, in my opinion a really good bull. He is a proven calving ease bull but his sire Grid Maker, while a good bull in own right for growth definitely is not a bull you'd want to breed heifers to and expect problem free experience.
We got a Hereford bull sone years ago that was a fire and ice mating. Average EPD's for the bull, he was touted by the sales promoter as a sleep all night calving ease heifer bull because his sire was.
I didn't think he looked like what I would call a heifer bull, but we got in a pinch and didn't have another bull at the time. He had sired average sized calves from cows with no problem so we took a chance and put him with sone heifers. Had to pull over half of the calves and sone were very hard pulls.
He must have taken after his dams side similar but opposite of what Bismarck did.