I guess I could try and calibrate my readings based on the reflections
If you were using a laser there would not be much need for calibration, if the foil obscured the mirror the beam would be cut off.
Alternatively the mirror could be foil width and on the perfect track. If the beam gets 'made' the mirror has been exposed (I think I prefer it being cut).
The video showed an essentially continuous process. What I was trying to suggest is that perhaps a batch approach might be better if you are only rolling 1m lengths.
It sounds as if you have an existing machine though. What is the actual application, and what is the current machine design if that is not asking too much?