A lot has to do with the fact that AI sires are able to transmit the quality across a broad spectrum of cattle. In other words more often than not the calves out of said bull will be better than the dam they are out of but not necessarily better than the sire which is why very few sons of great AI sires become great AI sires as well, even though we know it happens all the time. Speculation on young sires causes the high purchase prices just like speculators in the oil industry.... it takes big risks to return big rewards and then sometimes you just get plain lucky.
Even though G A R Integrity sold for $255,000, and then proceded to flop majorly, the folks that bought him made money because everyone bit into the hype. However, Integrity wasn't a lasting force due to the lack of performance.
You can find several young sires as well as young females that are non-parents and have incredibly high EPD's for WW, YW, %IMF, UREA, and $B that will eventually fall into line with the masses due to their progeny. This happens due to how the animals where set up into contemporary groups, how they were fed within those groups and what the pedigrees of the other animals in the group were. If you don't think some ranches influence EPD's legitimately through selective contemporary grouping you are naive. However, when these animals leave their original herd their true value often starts to show.....especially with sires that are promoted for AI use.
So yes marketing has a lot to do with it but the accuracies of the EPD's has a bigger influence in the long run. As well as the longevity, phenotype and general production aspects that are not quantitative.