Hello Hector,
As one control expert said, "All process models are wrong but some are useful!" Howwever, the more accurate the process model, the better analysis. When we say a response is "first order self-regulating plus dead time", we really mean it fits this model reasonably well. When the process has additional "lags", a "second order self-regulating plus dead time" model is a better fit for the process. If you model this as a "first order plus dead time" process the extra lag(s) will cause an increase in the "apparent" dead time. Examples of multi-lag processes include temperature with heat transfer across a thermowell and heat exchanger surfaces; temperature in a multi-tray or packed distillation column; fired heater with heater transfer through tubes and a cascade loop, etc. If you use a "process model based tuning method", such as Lambda Tuning, the secondary lag or "time constant" is used to calculate the derivative setting should you choose to use derivative. Some other tuning methods don't measure this time constant but rather use rule of thumb to determine the derivative setting. Regardless of whether you use derivative or not, have a more accurate process model, including the seconday lag if appropriate, will provide a more accurate anlaysis of the performance of the selected tuning.
Hope this helps!
James Beall