Hi Juyong,
The question you have asked has already been answered at this community post. To summarise the reason for the difference – When you call the function with 3 return variables: [p, S, mu], the third output mu tells matlab to center and scale the data. The centering and scaling values are stored in mu. This centered and scaled data is used to generate the polynomial fit. The polynomial fit in the function call without mu is created using the raw R1 data. Hence, there is a difference between the parameter values or p vector in the two cases.
You can read more about the polynomial fit function at the documentation page
