Volume/Gain for re-recording from my interface

Can anyone school me - I am going nuts trying to understand where I am going wrong trying to do what I thought should be a simple test:
 
I want to test my audio interface fidelity by taking an imported ripped wave from a CD and sending it out a stereo pair on my interface while recording it back in through a stereo pair.  The idea is that I'll repeat this 10x or 20x and see what kind of degradation and artifacts I notice when I A/B the first generation and last generation recordings.
 
Here is what I did:
 
1. I ripped a lossless wav file from a CD to serve as my source audio
2. Opened Sonar and dragged the .wav file into a track.
3. Configured the output to ports 3/4 on my interface
4. Wired 3/4 out to 3/4 in
5. Added another track, set it to record from 3/4 in
6. Recorded a second pass from track1 to track2
 
What I expected was that I would see track2 matching track1, at least with respect to levels.  All faders and gain are at unity.  I use a Focusrite Liquid56 and as far as I can tell, gain is unity.
 
What I actually see is that the recorded track is hotter than the source track by a seemingly variable amount.  I've tried changing the volume on my source track1 to -1.0db and -0.5db and while this seems to give me a similar level on my track2, when I play back track2 to repeat the process to track3 for a third generation, the volume is mismatched again.
 
I can't help but think I'm doing something wrong ...  I've been unable to record several generations and have the volume matched, and I don't understand why.
 
Have I overlooked something simple?
 

There is nothing within Sonar that lets you modify the volume of a signal being recorded.
 
Any perceived change MUST be attributable to something outside Sonar in your recording chain.

Cactus Music
You test is pointless. Setting controls at unity does not mean the signal will not be changed. 
 

The op problem actually suggests that there is in fact a level descrepency when he takes his tracks out of the box and back in.  On many interfaces finding unity gain is not crystal clear, so this trouble shooting technique may help clarify whether its a coloring from the liquid preamp, thats causing the descrepency, or actual voltage.. i.e. if hes using the mic level instead of line etc.

Cactus Music
You test is pointless. Setting controls at unity does not mean the signal will not be changed. 
 

 
I might not have explained well, but I don't think it is pointless, and it is not about gain, per se.  What I am testing is the artifacts of my AD/DA conversion stages and analog circuitry by running a track out a pair of DACs and back in another pair of inputs though the ADCs, through several generations to enhance the artifacts.
 
Where I am confused is that I cannot seem to easily match levels and I expect if I record something in and it registers at -3.0db that I would get -3.0db on the analog outputs if I set the Sonar track volume to 0.0db and gain at 0.0db and have my interface at 0.0db (unity) as well.  In reality, I seem to be seeing shift, and I'm having a hard time compensating for it so that each generation has the same levels.
 

brundlefly
It occurs to me that another source of variation from one recording of the same source to the next will be the performance of the analog pre-amps and anti-aliasing filters on the inputs. It's doubtful they will behave exactly the same way every time. That's a lot of what your serial re-recording test is going to be measuring.

Ah - it may be this.  I had not considered anti-aliasing filtering either, and it might account for a small loss.  However, I think after watching the video linked above that I made a mistake in how I set CueMix to get unity gain, so I have an offset.  I'm going to try it all again. 

An individual re-recording may not be anywhere near unity gain, depending on how well calibrated the the analog output level and input sensitivity of the interface is, and whether they are both calibrated to -10dBu or +4dBV line level (not necessarily the case). There should be controls on at least one side (output or input) to allow you to match them.
 
At that point, you should get relatively repeatable recordings of a fixed source but you'll still have small, random variations in peak dB readings due to the inconsistency of the analog circuitry and the relative timing of the sampling, above and beyond the generational loss of 'fidelity' you're wanting to investigate.