If Cakewalk doesn't view it as a bug, you can enter it as a Feature Request:
 
http://www.cakewalk.com/support/contact/featurerequest.aspx
 
But, as I said previously, I don't view this as a bug, either. If a controller is sending Note Ons and Offs so close together that they get clocked in on the same tick by the MIDI interface, that's not really SONAR's problem. Yes, SONAR could be programmed to convert zero durations to some non-zero value, but rather than have every MIDI sequencer application developer do this individually, it would be much better for the controller manufacturer to program in a minimum duration between sending Note On and Off messages.

which drum pad are you using?  
edit - just watched the video... sdmp1... 
the first thing that jumps out is that you're quantizing only note length - which will make the note end on the next eighth note - the reason your durations are over 480 for those notes.
 
the other things i would look at are:
on your midi track, are you specifying midi channel 10 in the input from your controller or is it set to omni??
is your midi clock sync on the sdmp1 set to master or external?  if its master, set it to ext.
make sure soft thru is turned off on the unit.
try one of the other velocity curves in the unit and see if the results are the same.
 
the sdmp1 does send a note on message with a velocity of zero as a note off, so that's not the issue.