I actually do think they are valuable in the limited scope they were intended when the accuracies get at least above the genetic heritability of the trait . I think they are more valuable when the accuracies are above .90. Rarely do I see or hear the question asked for example, (which of these too bulls would weigh more at weaning) . To me the question goes more like which of these two bulls will reach my goal (ie move my weaning weight from 550 to 650 lbs.) There is no epd that can tell you the answer to that question.
I guess to me they answer a question that doesn't have much value to it if you know the answer.