I won't help directly because of my immediate absence from the collab. Your work sheet does not download, extremely rare. Richard repost is all scrap to the point that I can't even guess if there is something in it. XFR [Transfer Function] has a unique meaning in control technology. It results as a pure function from the system, therefore there is nothing to fit. If you have a data set that you believe is a XFR, it is not unless extracted from an XFR itself. If properly collected open loop, you have the "Reaction Curve". That both are very alike: yes. If you would want the best controlling parameters, you can derive them from Ziggler_Nichols and the reaction curve. What you could do is try different XFR from models and look for some best fit to the data set [Reaction curve]. At this point and if happy, you could them resume the system from the "accepted XFR". At this point then, you would just plug the project in the Mathcad qs "PID" and go for the best controlling scheme.
There is a work sheet that will "best fit Odesolve" a physical model. But, not seeing what it's all about, it would surely confuse.
jmG