There are two principal models that FlowJo can compute for any given distribution: the "Watson Pragmatic" model and the "Dean-Jett-Fox" model.
The Watson model makes no assumptions about the shape of the S-Phase distribution; it (by definition) fits the S-phase exactly. The DJF model assumes that the S-phase is can be modeled by a second degree polynomial (that is convoluted with gaussian distributions of varying width throughout S-phase). You can also choose to have a synchronized peak within the S-phase as an option to this model.

Once you select the model, you can also choose to manually edit the curves to better fit the data. Each gaussian peak (G1 or G2) can be edited by choosing 'Edit Model Curves'. This allows you to drag and resize the gaussian curves, and to view the recalculated statistics.

More hints on fitting the data can be found on the next page.