Lean operation is beneficial to spark-ignition engines due to the high thermal efficiency compared with conventional stoichiometric operation. Lean combustion can be significantly stabilized by the partial fuel stratification (PFS) strategy, in which a small amount of pilot injection is applied near the spark energizing timing in addition to main injections during intake. Furthermore, mixed-mode combustion, which makes use of end-gas autoignition following conventional deflagration-based combustion, can be further utilized to speed up the overall combustion. In this study, PFS-assisted mixed-mode combustion in a lean-burn direct injection spark-ignition (DISI) engine is numerically investigated using multi-cycle large eddy simulation (LES). To accurately represent the pilot injection characteristics, experimentally-derived spray morphology parameters are employed for spray modeling. A previously developed hybrid G-equation/well-stirred reactor model is extended to PFS conditions, to capture interactions of pilot injection, turbulent flame propagation and end-gas autoignition. The LES-based engine model is compared with Reynolds-averaged Navier-Stokes (RANS) based model, allowing an investigation of both mean and cycle-to-cycle variation (CCV) of combustion characteristics. Instantaneous spray and flame structures from simulations are compared with experiments. The LES-based model is finally leveraged to investigate impacts of fuel properties including heat of vaporization (HoV) and laminar flame speed (SL). It is shown that overall, the predicted mean pressure and heat release rate traces from both RANS and LES agree well with the experiment, while LES captures the CCV and the combustion phasing in the mass burned space much better than RANS. Predicted liquid fuel penetrations agree reasonably well with the experiment, both for RANS and LES. Detailed flame structures in the simulations also reveal the transition from a sooting flame to a lean premixed flame, which is consistent with experimental findings. LES is shown to capture more wrinkled and stretched flame fronts than RANS. Local sensitivity analysis further identifies the stronger combustion phasing sensitivity to SL compared with that to HoV, and the stronger sensitivity of autoignition heat release rate than deflagration. The results from this study demonstrate the high fidelity of the developed computational model based on LES, enabling future investigation of PFS-assisted mixed-mode combustion for different fuels and a wider range of operating conditions.