From the "unknown fake data set" we can't decide if the phenomenon
is really a two segments. In "Norman language" maybe YES, maybe NO ?
At least, Smath can do a great job for your friends in that business.
In the Mathsoft Collaboratory, this project was exhausted in two more
ways to filter the "assumed" raw data set; Spline2XV6 and Gaussian
b-spline. My "spy tool" confirms the data set is two badly made segments.

Jean

Genfit Oxyhemoglobin Reconciliation.sm (47.57 KiB) downloaded 800 time(s).