Why is there an Audio Bit Depth when Splitting Channels

I am working on a client's audio files which he sends me in stereo. The left channel is one person, the right channel is another person. This is a podcast and the right channel is a person on a telephone line and my job is to "enhance" the audio of the right channel only.
 
I imported the audio file into Sonar 8.5 and then bounced to track choosing the "split mono" channel format to separate the channels.
 
The original file is 48 / 16 bit. I noticed in the clip properties of the bounced file, the new tracks are 32 bit (also verified by Project / Audio files).
 
Why the bit depth change? Will I have to dither these files back to 48 / 16 for the client?
 
(As an aside, the new files are ~ 3db louder but I think I understand that. It also occurs when changing the interleave button on drum tracks from stereo to mono.)
 
Thanks! RK

OK, I see. Thanks!
 
Well then the question becomes, should I leave it at the default of 32?
 
From Help file:
 
Render Bit Depth. Use this field to choose the bit depth that audio processing (freezing, bouncing, applying effects) will use. The default value of 32 allows virtually lossless processing.
 
So should I render at 32 and dither to 16 upon export. Or render at 16 and export with no dithering?

I think I'll leave it at the default 32 bits to get the lossless processing and then use the ozone high quality dithering to reduce the file to 16 bits.
 
Thanks for the kind direction, it's a big help and now I don't have to worry about that; back to the task at hand!

Good point, thanks :)