Currently available density functionals cannot describe the dispersion component of the interaction energy present in weakly bound complexes. Moreover, the exchange energy as obtained from the density-functional theory is often incorrect. Examples of problematic cases include clusters of van der Waals-bound rare-gas atoms and most hydrogen-bonded molecular systems. Thus, accurate ab initio methods to treat intermolecular forces should be used in such systems. These methods are, however, too slow to be applicable to the large systems needed to model adsorption. This is why DFT continues to be used, where, in addition, a quite common compensation of errors sometimes produces some sort of agreement with the corresponding experimental data. In this paper, we analyze in detail the inadequacy of standard DFT for describing the weak binding present in a few rare gas-rare gas, metal atom-rare gas, and metal atom-metal atom dimers. Inspired by the success of the Hartree-Fock plus (damped) dispersion (HFD) method, we test the use of an improved hybrid model in which to a density-functional interaction energy (with corrected exchange and avoidance of double-counting of dispersion), a (damped) dispersion expansion is added in the usual way. Comparisons with accurate theoretical or experimental benchmarks show that our DFdD method using the revPBEx or revPBEx+VWNc functionals and accurate dispersion coefficients is found to recover the interaction energy curves very well for many of the tested systems. The second paper in this series will describe the use of the DFdD method for physisorption for the previously well-studied (but not solved) case of Xe/Cu(111).