Years ago zubinsum posted a list of pitchers that seemed to be under and over priced base don the current base model for pricing. And I pointed this out back then, and still hold to it (especially post testing), that they weren't underpriced or overpriced, but his valuation wasn't including the errors from normalization or the pitchers hitting stats.
Same issue arises in cookies, people gravitate to the deadballers because they LOOK like a bargain when they have better WHIP, OAV, and ERA, which are all positively impacted for the pitcher with errors. Errors reduce their OAV, their ERA, and have no impact on their WHIP. So, the pitchers that had more errors behind them look better and when using them in sim, the stats do the same thing. They see Bernhard put up a 3.20 ERA and think he was great, but the errors he allowed had him giving up 5.20 R/G, and the '76 Blue will put up similar R/G with an ERA of 4.05 due to fewer errors and people think at a glance that Bernhard was the better pitcher, when they were essentially identical.
It's why '08 Joss was so popular. He looked to be a great buy at a glance, but ultimately his performance wasn't any different than they other $12m pitchers when you factored errors in. I build most of my teams around the idea that most of my opponents will have deadball pitching and guys like Bernhard often give up 6+ runs against my teams due to the errors.