The third-order reaction, H + O-2 + M --> HO2 + M, has been measured near the low-pressure limit at room temperature for M = He, Ne, Ar, Kr, O-2, N-2, and H2O and over an extended range of temperatures in a shock tube for M = Ar, O-2, and N-2. In all cases, H atoms were produced by the laser photolysis of NH3 and detected by atomic resonance absorption spectroscopy. The measurements are consistent with the available experimental record and, in particular, confirm the exceptionally high recombination rate constant when M = H2O. The standard theoretical analysis is applied to this entire experimental record to derive the value of the average energy change per collision, -DeltaE(all). The resulting -DeltaE(all) values are sensible for all M but H2O. The problem with H2O motivates a change in the standard theoretical analysis that both rationalizes the behavior of H2O and also quantitatively changes the derived -DeltaEa(ll) values for the other species of M. These changes involve three modifications of the standard treatment: (1) explicit temperature dependence in the number of active rotational degrees of freedom contributing to the HO2* state density, (2) the replacement of Lennard-Jones potential for the HO2* + M interaction with an electrostatic + dispersion potential, and (3) the calculation of the collision rate between HO2* + M by a free rotor model for "complex formation" between the M and HO2*. The optimized values of -DeltaE(all) that are produced from this new analysis have the following characteristics: (1) the value of -Delta(all) is the same for all rare gases, and (2) -DeltaE(all) for di- and polyatomic molecules are enhanced relative to the rare gas atoms. This work supports the conclusions of previous trajectory studies that collision rates between activated complexes and bath gases are often underestimated while -DeltaE(all) derived from recombination kinetics measurements are often overestimated.