An integrated structural, petrological and geochronological study was undertaken to constrain the tectonic history and controls on the exhumation of the high-P rocks of the Río San Juan complex in the northern Caribbean subduction–accretionary wedge. In the main structural units of the complex, microtextural analyses were performed to identify the fabrics formed at peak of metamorphism in eclogite-facies conditions and during the main retrogressive event toward the low-P amphibolite or blueschist/greenschist-facies conditions.U–Pb SHRIMP dating on zircon rims (71.3±0.7Ma) coupled with 40Ar–39Ar analyses on phengite (~70–69Ma) in felsic sills placed temporal constraints on the exhumation of the Jagua Clara serpentinite-matrix mélange during the blueschist-facies stage at the early Maastrichtian. In the Cuaba unit, U–Pb TIMS zircon ages of 89.7±0.1Ma and 90.1±0.2Ma obtained for the crystallization of tonalitic/trondhjemitic melts in the lower Guaconejo and upper Jobito subunits, respectively, are similar. These ages coupled with a U–Pb SHRIMP zircon age of 87±1.8Ma obtained in a garnet amphibolite and a group of older 40Ar–39Ar cooling ages on calcic amphibole constrain the exhumation of the Guaconejo subunit from the high-P stage to the low-P stage at the ~90–83Ma time interval. Further, the age data collectively supports a genetic relationship between the distributed extensional ductile shearing, the related decompression and the local partial anatexis in the subunit, at least from the Turonian–Coniacian boundary to the early Campanian. A group of younger 40Ar–39Ar ages obtained in the mylonitized amphibolites of the basal Jobito detachment zone indicates late ductile deformation and exhumation/cooling in the late Campanian to Maastrichtian (~75–70Ma). Therefore, structural and age data established deformation partitioning and reworking of retrograde fabrics during ~20Ma in the Cuaba unit.The different exhumation rates obtained for the Jagua Clara mélange can be explained by uplift in two contrasting tectonic settings: a first stage of slow exhumation (1.4mm/yr) in the subduction channel, largely lower than plate velocities, and a second stage of relatively fast exhumation (7.6mm/yr) up to the surface. Therefore, the exhumation was temporally discontinuous and the velocity increase at ~70Ma probably was triggered in response to the entrance of buoyant material in the subduction zone, such as a Caribeana continental ribbon or the distal part of Yucatán–Bahamas continental margin. In contrast, the exhumation path of the Guaconejo subunit is composed of a first segment from the baric peak to the low-P amphibolite stage (at 84–83Ma) with an exhumation rate of 7.2mm/yr, a second segment around the closure temperature of calcic-amphibole (at 82–70Ma) with a rate of 0.4mm/yr, and a third segment to the surface exposure (at 60Ma) with a rate of 1.8mm/yr. These velocity differences can be correlated with the P–T path proposed for the exhumation of the subunit, with an initial isothermal decompression from the peak in the high-P amphibolite to the eclogite-facies produced by distributed extensional shearing, followed by a relatively slow cooling at low-P in the newly acquired structural position, and the final tectonics mainly partitioned in the late Jobito detachment zone. The initial fast exhumation can be related to a major modification in the convergence conditions across the intra-oceanic subduction zone as a jump of the basal subduction thrust beyond a previously accreted arc terrane.