Here's the thing - you're going to rate the Castles as a single pile all together, right? It would be silly to rank them all individually. But the same logic you're using on Split piles can be used on them. They all have different costs, you don't need to buy one to get another one, and there are even multiples of some of them with more than 2 players! While we're at it, why don't we rank all of the Knights individually as well, putting Sir Martin on the $4 list for good measure?
I've had difficulty forming an opinion here. I started writing a response, and now I think I have a side here.
Split piles? Well...I feel that the existence of the bottom cards is a feature of the top cards. You rate Catapult knowing you can get Rocks. You rate Encampment knowing you can save them with Plunder. Similar story with Patrician and Gladiator (remember you gain Gold with Fortune for each Gladiator you have in play, and you can help reveal Fortune with Gladiator's on play).
So, I personally prefer the strength of the bottom card adding to the overall value of the top card and have then rated together as one card pile. Sure, you can get the bottom cards without getting the top cards, but you can see that as a sort of penalty your opponent gets for emptying the top card and factor that into the rating of the top card.
You may never get to the bottom cards in a game, but like sometimes you never get to play an activated City. The potential to activate City is still relevant to the strength of City.
You can rate the two cards in the split pile separately if you really want, but I'm left wondering if you really gain much from doing that. I think a unified ranking gives a better sense of how the pile will impact the game.