The data sampling is limited by the pressure analyzer output time delay. It only analyze the sample every 15-20 min. The results are not like a pressure transmitter, they differ from one reading to another.
All of the data is taken when the master loop (pressure) is in manual and flow loop (slave in cascade). You can see as the pressure loop output (0-5%) increases the pressure loop pv (kpa) increases as well.
The goal here is not to be exactly at the pressure setpoint but to be near it. Under these conditions can a 1st order response be trusted from PID tuner?