Sorry for the delay. I've been on travel.
For the second part:
1. When you are using an asset audio player, are you using the same content as for the buffer queue? From your description I believe that you have a file that you decode to PCM not through the OpenSL ES implementation and then submit it using the buffer queue. You are then getting the wanted result from the reverb. That is good. Then you play the original file through the system codec and using the same code except that you use a SLDataLocator_URI instead of SLDataLocator_BufferQueue. I can't explain it, but you should be getting the same results if you use the same input. Have you tried to load the PCM data through a file?
2. Yes and no. OpenSL ES was not really designed to do this - one reason is to protect DRM protected content. But *if* the system allows it, you can use a buffer queue as a data sink for the player. Support for this wasn't added until 1.1 when we added object chaining so that you can hook a recorder to a player object in order to do transcoding.
For your transcoding the MP3 into PCM and submitting it through a buffer queue, have you verified the pcm data using other means? Since you have successfully managed to submit through the buffer queue then you know the chain works. If you are getting noise then I would suspect your PCM data.