Microwave (wave ~1.2 mm) spectral line observations of HO2, H218O, and O3 were obtained at the Kitt Peak National Radio Astronomy Observatory (NRAO) on April 10--12, 1992. The spectral bandwidth and resolution of the observed collisional line shapes are appropriate to retrieving profile abundances for O3, H2O, and HO2 in the upper stratosphere and mesosphere (45- to 70-km altitude) with 20--30% uncertainties. The derived profile for O3 exhibits 20--30% larger mesospheric ozone abundances than indicated by the April 30¿N average profiles from Solar Mesosphere Explorer (SME) and Stratosphere Aerosol Gas Experiment (SAGE) II and 60-80% larger than a photochemical model employing current photochemical rate constants and an H2O profile constrained by the microwave H218O observations. Surprisingly, the microwave measurement of HO2 yields mesospheric HO2 abundances roughly twice those predicted by the standard model photochemistry. As HOx (OH, HO2, H) is the primary agent for catalytic destruction of ozone in the mesosphere, the model underpredictions of both HO2 and O3 suggest a change in the model photochemistry that is distinct from errors in Ox production and HOx loss rates. Increases in Ox production and/or increases in HOx loss rates have been considered as the likely sources for the persistent data-model disagreement in upper stratospheric/mesospheric ozone abundances. Based on the microwave measurements of O3 and HO2, we propose changes to model rates for partitioning between OH and HO2. Specifically, a large (60--80%) decrease in the rate coefficient for the reaction O+HO2→OH+O2 is shown to be uniquely capable of increasing model abundances for O3 and HO2, in accordance with the microwave observations as well as the outstanding underprediction of upper stratospheric/mesospheric ozone by photo-chemical modeling.