We have used an extended version of the Cascade-Exciton Model (CEM) to analyze more than 600 excitation functions for proton induced reactions on 19 targets ranging from 12 C to 197 Au, for incident energies ranging from 10 MeV to 5 GeV. We have compared the calculations to available data, to calculations using approximately two dozen other models, and to predictions of several phenomenological systematics. We present here our conclusions concerning the relative roles of different reaction mechanisms in the production of specific final nuclides. We comment on the strengths and weaknesses of the CEM and suggest possible further improvements to the CEM and to other models. * On leave of absence from Bogoliubov Laboratory of Theoretical Physics, Joint Institute for Nuclear Research, Dubna, RussiaPrecise nuclear data on excitation functions for reactions induced by nucleons in the energy range up to several GeV are of great importance both for fundamental nuclear physics and for many applications. Such data are necessary to understand the mechanisms of nuclear reactions, to study the change of properties of nuclei with increasing excitation energy, and to study the effects of nuclear matter on the properties of hadrons and their interactions. Excitation functions are more sensitive to the detailed mechanisms of nuclear reactions than are double differential cross sections of emitted particles or their integrals over energy and/or angles. Therefore, excitation functions are a convenient tool to test models of nuclear reactions.Second, and perhaps more important today, expanded nuclear data bases in this intermediate energy range are required for several important applications. Recently, one of the most challenging problems requiring reliable nuclear data files is Accelerator-Driven Transmutation Technology (ADTT) for elimination of nuclear waste [1]. The problems of Accelerator Transmutation of Waste (ATW) are closely connected with Accelerator-Based Conversion (ABC) [2] aimed to complete the destruction of weapons plutonium, and with Accelerator-Driven Energy Production (ADEP) [3] which proposes to derive fission energy from thorium with concurrent destruction of the long-lived waste and without the production of weapons-usable material, though substantial differences among these systems do exist [2]. Precise nuclear data are needed for solving problems of radiation damage to microelectronic devices [4] and not only of radiation protection of cosmonauts and aviators or workers at nuclear installations, but also to estimate the radiological impact of radionuclides such as 39 Ar arising from the operation of fusion reactors or high-energy accelerators and the population dose from such radionuclides retained in the atmosphere so as to avoid possible problems of radiation health effects for the whole population (see, e.g. [5]). Another important new application which requires large nuclear data libraries at energies up to several hundreds of MeV is the radiation transport simulation of cancer radiotherapy used f...