I was actually in the middle of typing that same information to you when I got a call! Sorry that I didn't identify this problem sooner, I misread the specifications sheet and didn't recognize the overvoltage Channel-to-Com voltage as being +/- 30 until today. This would definitely explain why were are seeing some funny behavior on the Line voltage.
National Instruments | Applications Engineer | www.ni.com/support
This forum, and particularly the schematic and screen shots, are eerily familiar. We are seeing the exact same results. So, I don't really anticipate that your answer will be much different, although I will say that you never know until you ask...
Could it be that reading in the NI-9215 spec to say that it's input range and capability are all with respect to (an assumed) ground potential is incorrect? Afterall, while the Operators Manual does say "Input range...+/-10.0V" in one area, it clarifies in the "Maximum Voltage (Signal + Common Mode)" section that this is "...+/-10.2V of common," and while the operators guide says "Overvoltage protection...+/-30V", in the "Safety" section this is clarified to be "Channel-to-COM...+/-30V max".
So, just for kicks sake, is it possible to get you to do a double-check and verify that the intended design is not to allow tying COM to AI- (and, for the sake of simplicity, let us assume that we have only a single channel implemented) and allow a true differential input signal over a range that keeps the device within its specified voltage safety limits?
I know - it's not working with this interpretation - but that doesn't necessarily validate that that is not its intended function.