That seems like a great benchmark to me. If trial and error gives feedback that allows you to qiuckly learn the UI then that's pretty useful.
At the end of the day a lot of usability is about two things:
1. Consistency 2. Familiarity
If trial and error is effective then consistency is in place. Skeumorphism is about taking familiarity of the real world and applying it to the digital world. And this actually has some use, in particular for UI that users won't interact with much. There's no chance for users to be become familiar with that UI. But for UI that is always there, skeumorphism becomes limiting and can become unusable when you really want to extend beyond what you see in the real world.
Metro says, "The digital world is becoming so prevelant that we should optimize for it, not just the physical world." Is Metro perfect? No. But I think it has the right idea to figure out what works on computers/tablets/phones first. Don't be encumbered by trying to map to physical objects. People will spend so much time with the digital objects that they may spend 10m doing trial and error the first time, but shortly the digital world will be just as familiar as the physical world. Lets not waste the opportunity to introduce the right interactions.