The 11 March 2011 Mw9.0 Tohoku-Oki earthquake was recorded by an exceptionally large amount of diverse data offering a unique opportunity to investigate the details of this major megathrust rupture. Many studies have taken advantage of the very dense Japanese onland strong motion, broadband, and continuous GPS networks in this sense. But resolution tests and the variability in the proposed solutions have highlighted the difficulty to uniquely resolve the slip distribution from these networks, relatively distant from the source region, and with limited azimuthal coverage. In this context, we present a finite fault slip joint inversion including an extended amount of complementary data (teleseismic, strong motion, high-rate GPS, static GPS, seafloor geodesy, and tsunami records) in an attempt to reconcile them into a single better resolved model. The inversion reveals a patchy slip distribution with large slip (up to 64 m) mostly located updip of the hypocenter and near the trench. We observe that most slip is imaged in a region where almost no earthquake was recorded before the main shock and around which intense interplate seismicity is observed afterward. At a smaller scale, the largest slip pattern is imaged just updip of an important normal fault coseismically activated. This normal fault has been shown to be the mark of very low dynamic friction allowing extremely large slip to propagate up to the free surface. The spatial relationship between this normal fault and our slip distribution strengthens its key role in the rupture process of the Tohoku-Oki earthquake.