If songs are being recognised on the N9 (but not as reliably as on the N950) perhaps the microphone is simply less sensitive in its default state. I'll try putting together a new version when I have some time at the weekend that boosts the input volume and see if that improves matters for anyone.

As to the number of songs available for identification that's a known issue with the EchoPrint database, although EchoNest are planning on expanding this in the near future.