Starship's limited $\Delta V$ means that it can't stray too far from minimum energy transfers.
Here is a porkchop plot of the next decades' worth of Mars to Earth trajectories subject to $\Delta V < 6.9$ $km/s$ (ignoring fuel for landing) from a low Mars orbit (150 km), notice the sparsity above the black 'equal date' line:

Those are some ludicrous entry speeds at the high end of the scale, though they are uncommon. If we constrain the entry velocity (inertial) to 12.9 km/s, the highest ever for something human made (Stardust, 2006), we get something like this in the 2028-29 range:

The minimum $V_{Entry}$ for the decade is $11.4$ $km/s$, though given some of the more efficient (~$2$ $km/s$ $\Delta V$) transfers it's theoretically possible to perform a pre-entry 'braking burn' with the excess fuel remaining to reduce entry velocity significantly.
Crew Dragon uses PICA-X, a SpaceX derivative (& improvement) of PICA used by Stardust. It is an ablative heatshield material. I suspect the current Starship uses some kind of PICA-X like material given this statement on SpaceX's Starship page:
Starship will enter Mars’ atmosphere at 7.5 kilometers per second and decelerate aerodynamically. The vehicle’s heat shield is designed to withstand multiple entries, but given that the vehicle is coming into Mars' atmosphere so hot, we still expect to see some ablation of the heat shield (similar to wear and tear on a brake pad) [...]
Given Stardust's success (very cool video of Stardust entering the atmosphere), it's probably safe to suspect that Starship's heatshield could handle a direct entry, if only once.
Aerobraking is a difficult problem for the ablative heatshield. As explored in my answer to What is the heat shield refurbishment procedure for a crew Dragon capsule? PICA(-X) is highly capable, but the underlying mechanism of ablation does not lend it self to rapid reuse (i.e., repeated aerobraking passes). Here is a series of plots detailing three trajectories (all $11.4$ $km/s$ initial entry velocity): direct, ~lunar distance aerocapture, 5 G's max aerocapture along with a LEO reference.

The top row plots show the flight trajectories and the bottom row of plots can be interpreted as 'heat shield strain' as explained in my linked answer (heat flux VS stagnation pressure).
It can be seen that both high (instantaneous) heating and high deceleration can be avoided using aerobraking, though the integrated heat load, a strong determinant of heat shield thickness, remains more or less the same (~$200$ $kJ/cm^2$) for each Mars return architecture, ~16 times greater than a LEO return.