Robert, one more thing. What does your detector look like? I used to have the transmission densitometer standard around here, but I can't find it. Anyway, in that standard it defines how the negative shall be illuminated and the light from the sample is measured (the angle). Basically, if the light source is collimated, as through an aperture, the sensor should read as much of the resulting light coming from the negative, i.e. the sensor should have an 'integrating sphere' design. If the light source is diffuse, then the sensor should 'see' only a narrow angle. All angles on one side, narrow angle on the other.
A different density results if this is not followed, because of the Callier effect. This difference only matters if you want to be able to compare numbers with those in books.
It sounds like your light is collimated, therefore your sensor should have an integrating sphere or some kind of integrating quality. It needs to be able to see light coming from all angles of the negative. If the negative is in direct contact with the sensor, this works. Or if you put a piece of white Plexiglas in contact with the negative and then the sensor looks at the back of the white Plexiglas, this works too. Unfortunately, the Plexiglas attenuates the signal.
You could turn this around and do it on top of a light table. Mark a small area on the light table, put the negative there and put the sensor on top looking down. This is good if your sensor is not an integrator and sees only a narrow angle.
Hope this makes sense. Maybe someone out there has the standard and can correct any inaccuracies.