My take on this, is that the absolute best place for the audio to be decoded is in the Reciever/Decoder, and it will be perfect if the Reciever also has a Digital Output stage like TactAudio or some the highend Sony Recievers from the last years.
Feeding PCM to a reciever is the next best thing, it just requires the player to have all the needed controls for setting it up perfectly for your room and display.
Something like Room Correction, even on a very crude level is also very handy. But few, if any player will support this.
So the problem is that eventhough the audio signal is in a digital domain (being PCM( the handling of it in the reciever is hindered by the lack of information that was in the bitstream, if it has to process to PCM stream in any form it may result in resampling or just generally screwing around with it. And since the encoding information is gone from the PCM streams at best it can only some more general stuff you can do to the stream, again at the risk of making it sound worse.