The API copyright is but one part of the whole trial. Even after the jury comes back, there is still a patent issue and then damages portion (though this depends on how the jury decides in the prior two phases). And IIRC, the API copyright issue is but one part of the copyright complaints brought by Google (though I think most/all of the others were already tossed).
Judges do things like this a lot. If the Oracle legal team presented what he thought to be a good case in favor of the copyrightability of the APIs, then he might've decided to let it go to the jury rather than let Oracle appeal. This way, Oracle can't say they lost because they couldn't present their argument, and the judge can use case law later on so that Google can't appeal because the jury had no clue what they were talking about.
Finally, just because case law has set a precedent does not necessarily mean that the precedent is correct or that a future case can't lead to that precedent being overturned. This is in large part why our system exists as it does, with courts of increasing national authority that can step in and correct a lower court for decisions which should not have been rendered or for the abdication of due process.
Understand, I certainly don't want Oracle to win this one, but I do understand the judge's thinking. This isn't an inefficiency of the judge, it's the judge exploiting his knowledge of the system he works in every day.