Hi Andrew, it seems like a relatively complicated setup with many variables (ie potential sources of error) to control. Here are some additional things to consider:
-Reciprocity failure: Assuming I'm following your methodology correctly, reciprocity failure will have a greater influence as exposure time increases (since you are doing separate exposures). This could help explain the flattening of the highlights. It is best to eliminate reciprocity failure from the test. Using an enlarger as the light source for an in-camera test is problematic because of the relatively low level of illumination.
-Lens extension: For an in-camera test it would be best to focus the lens at infinity. Focusing on your close target means the lens is likely extended enough to require exposure compensation versus the handheld meter.
-Half-stop resolution: If you have limited materials, and given the variables involved, I'd suggest doing the test in full stops to simplify things. You can interpolate between points. Alternatively - if you contact a step tablet you could plot the full curve in 1/2 stops or 1/3 stops using only 2 or three frames of film. It's not an in-camera test, but it eliminates many variables, including flare.