Despite continuous improvements in seismic acquisition and processing technology, imaging under salt remains challenging, specifically because of the difficulty in updating complex salt geometries and subsalt velocity. Synthetic studies show that when certain conditions are met, full-waveform inversion (FWI) can recover very complex velocity models, including the geometry of the salt and the subsalt velocity. Unfortunately, currently available seismic field data fall short of meeting the requirements needed to replicate what can be achieved on synthetic data. We first use a wide-azimuth data set from the Mexican side of the Gulf of Mexico (GOM) to show how FWI can improve imaging in the subsalt. In addition to utilizing the diving-wave energy to derive a reliable model in the shallow sediment overburden, we use reflection FWI (RFWI) to update the velocity model in the deep area. RFWI utilizes the low-wavenumber components of the FWI gradient associated with waves reflected in the model, which makes it possible to circumvent the well-known penetration-depth limitation of FWI and the shortcomings of traditional tomography-based methods. This is achieved by alternately using the high-wavenumber and low-wavenumber components of the FWI gradient to update density and velocity models, respectively. We then use an ultralong-offset, full-azimuth data set from the U.S. side of the GOM to show that, with more suitable data, FWI and RFWI can be combined to recover the velocity in and around complex salt bodies, providing significant uplift to subsalt images.