I think it's WAY more important to make it possible to play sounds by index as with the native XACT, thereby making it unnecessary for us to keep a bunch of strings in memory.
I currently have a large string table because of this that I would like to eliminate.
There are even more missing features that the native XACT has that should be implemented like notifications which are quite handy when preparing streaming wavebanks. Why not utilize .Net events for this? Seems like a good candidate. As it sits, we have no choice but to implement a thread or series of non-blocking calls ourselves to constantly check the IsPrepared flag. In addition, I noticed there's no functionality for preparing the soundbank for zero-latency streaming...
Also, marker support as in the native libraries is needed. I really hope that these features are not overlooked because the ones being talked about above are going to be used in a very small minority of cases and these features I have mentioned are already supported and just need to be exposed.
I noticed the newer updates to XACT also have support for 3d panning and a better interface. It sure would be nice if the XNA GSE version wasn't so time lagged that we have to be stuck with an old version.