In principle, including multiple periods of the structure in the unit cell, hence making the unit cell bigger, would not change the phase vs. wavelength relationship.
However, this will inevitably increase the number of diffraction orders supported by the grating. The additional orders have near-zero strength, though. As a result, the total number of meaningful orders will stay the same.
I do not think you gain anything by Including multiple periods in the unit cell. It results in increase in memory requirement and simulation time. The results also include unnecessary zero-strength orders.
As to the phase of the unit cell, the results will be the same if you do some convergence test on both cases.
Here are examples of a grating with a single period and three periods:
To produce the above results, open the simulation file [ 30966_2.fsp (276.3 KB) ] and run the script [ 30966_2.lsf (692 Bytes) ].
If you are interested in the phase of the nearfield for the entire metalens, you cannot use the grating projection. Instead, you will need to use the nearfield result over the surface of the metalens. If the period of the unit cell is subwavelength, the injected will propagate mostly like a planewave. In such case, the phase directly from the nearfield would be close to the phase calculated from the grating projection.