Accurate and efficient estimation of aerothermodynamic loads is a fundamental challenge for multidisciplinary modeling and analysis of ultra high-speed vehicles. This study focuses on this issue by assessing the accuracy and studying the impact of combined theoretical and data-driven models over a broad operational space. A computational fluid dynamics (CFD) enriched piston theory (EPT) approach is benchmarked against both CFD and a basic engineering approach from a combined shock-expansion and third order piston theory (SEP) model. In the EPT approach, the enrichment is supplied by Kriging interpolation over a set of steady-state Reynolds-Averaged Navier-Stokes (RANS) solutions. The first part of this study focuses on offline assessment of the accuracy of the reduced models relative CFD. The second part carries out online comparisons of the impact the reduced models have on the dynamics of a free flying vehicle. The EPT approach is found to yield improved agreement with CFD relative to SEP for offline predictions. Furthermore, the use of EPT for online loads prediction of a free flying vehicle significantly shifts open loop vehicle response, with SEP predicting larger inclination and roll angle changes than EPT. Both aft body flow interactions and modeling errors are discussed as potential causes.