Probably the key term here is pork chop plot. The linked page includes one from 2005 for Mars where the lowest C3 departure was 15.5 and 400 or so days, but doubling that to 30 got a transfer in 125 days, by leaving Earth faster and burning off more energy on arrival.
For science payloads it generally makes sense to use the minimum fuel load and pack as much equipment as possible and wait the extra days.
For humans they need to eat and breath, so there is math to be done removing consumables and adding fuel. In the 2005 porkchop plot going from 15.5 C3 to 16 gets access to a new minimum on the lower part of the plot, dropping from 400 to 200 days, and pushing to 16.5 gets to 175 days. So for Humans the physically optimal transfer is probably never going to be the best choice, it just becomes a question of trade offs.
With SpaceX, having made Starship the all in one solution for Mars they have a heat shield that is (hopefully) robust enough to re-enter at Earth, and has enough performance to reach Earth orbit in it's own right (many Mars designs assume ion engines). So they have a design that is consumable/space constrained but has surplus performance for departure and capable of aerobrake/aerocapture at Mars, which makes high energy routing more possible than for a larger but lower thrust/less robust vehicle.
Possibly most critical Starship uses cryogenic Methane and Oxygen so every day in space loses some to evaporation. So Starship has further choices to burn fuel on departure for faster transit or have to carry more spare fuel to make up for more evaporation.
It is possible that the 90 day transit is being forced on them by evaporation rather than a true feature of the design. At the moment (January 2022 before first sub orbital Starship flight) not enough information is available for outsiders to apply useful math to the question.