Deep and fast wide-angle reflection arrivals observed at offsets over 180. km, and over a reduced time interval of 1-1.5. s, have been observed in a seismic experiment shot across SW Iberia as part of the IBERSEIS project. Using different modelling approaches, these reflections have been found to be consistent with a heterogeneous gradient zone located at 61-72. km depth that features an absolute P-wave velocity contrast from 8.2 to 8.3. km/s. Paradoxically, this interface has not been observed in coincident vertical incidence data, probably due to the change in the reflection coefficient with decreasing incidence angles, the lack of energy at high recording times for the near-vertical (vibroseis) data, and/or the different location of the CDPs in both experiments. Although the mantle is acknowledged to be highly heterogeneous and mantle lithologies are capable of giving impedance contrasts high enough as to be observed in seismic data, it is often seen as transparent from a seismic point of view. The short wavelength of mantle compositional heterogeneities is probably what hinders their identification with active source seismic data and only big and sharp discontinuities are imaged in vertical incidence experiments whereas regional transitional boundaries may be also observed at high incidence angles. Accordingly, we propose that deep reflectivity observed in SW Iberia must correspond to a regional-continental scale feature, not sharp enough as to be seen with vertical incidence energy. This feature, already identified in previous DSS experiments carried out in Iberia, has a depth, a P-wave velocity contrast and a transitional nature that match the characteristics proposed for the spinel-lherzolite to garnet-lherzolite phase transition, i.e. the Hales interface or gradient zone. This boundary is relatively narrow (at least 2-3. kb) in enriched mantle and appears deeper and along wider intervals when the mantle is depleted. In addition, it is a worldwide scale boundary already identified over large areas with different types of datasets. The variability in depth and sharpness of this interface, which is related to mantle chemistry, constrains the type of seismic techniques that should be used to identify it.