It is impossible if you're just an application. As an application you just create and start the buffers that need to be mixed. How and when the buffers are mixed (in software, hardware or even analog) is up to the driver of the soundcard. You can't capture the mixed result. Some soundcard drivers allow to route back the wave output to the record-in though.
However, I remember a proxy soundcard driver that tells the system he is just able to playback a singe sound stream. This tells DirectSound that it has to do all the dirty work (e.g. mixing). The driver had the feature to route that either to another soundards primary buffer or a wav-file. I don't remember how that driver was callled, but it's widely used by those folks who try to rip music from DRM secured sources. I'm sure if you really search for it you'll find it.
Another solution to your problem would be to write a sample mixer yourself (hint: if you haven't already done so, you'll learn a lot!). Depending on your skill that might take something between 2 days and a month. It's not that difficult and if you dig a bit into resampling you'll learn a lot about signal theory, what's involved when it comes to quality ect. Most drivers just do it the dirty way (the ms sample mixer is no exception here). The quality boost you can get from a proper implementation that doesn't need to work in realtime is incredible.
So, sorry - no quick solution to your problem, but some hints and ideas...