We outline our methods for obtaining high precision mass profiles, combining independent weak-lensing distortion, magnification, and strong-lensing measurements. For massive clusters the strong and weak lensing regimes contribute equal logarithmic coverage of the radial profile. The utility of high-quality data is limited by the cosmic noise from large scale structure along the line of sight. This noise is overcome when stacking clusters, as too are the effects of cluster asphericity and substructure, permitting a stringent test of theoretical models. We derive a mean radial mass profile of four similar mass clusters of high-quality HST and Subaru images, in the range R=40kpc/h to 2800kpc/h, where the inner radial boundary is sufficiently large to avoid smoothing from miscentering effects. The stacked mass profile is detected at 58-sigma significance over the entire radial range, with the contribution from the cosmic noise included. We show that the projected mass profile has a continuously steepening gradient out to beyond the virial radius, in remarkably good agreement with the standard Navarro-Frenk-White form predicted for the family of CDM-dominated halos in gravitational equilibrium. The central slope is constrained to lie in the range, -dln{rho}/dln{r}=0.89^{+0.27}_{-0.39}. The mean concentration is c_{vir}=7.68^{+0.42}_{-0.40} (at a mean virial mass 1.54^{+0.11}_{-0.10}times 10^{15} M_{sun}/h), which is high for relaxed, high-mass clusters, but consistent with LCDM when a sizable projection bias estimated from N-body simulations is considered. This possible tension will be more definitively explored with new cluster surveys, such as CLASH, LoCuSS, Subaru HSC, and XXM-XXL, to construct the c-M relation over a wider mass range.