A low Earth orbit (LEO) positioning, navigation, and timing (PNT) selection approach of satellites with poorly known ephemerides is developed via the dynamic data-driven application systems (DDDAS) paradigm. The approach integrates physical and simulated ranging measurements from visible LEO space vehicles (SVs) to select the best SVs to use. The selection is performed by minimizing the impact of SVs’ ephemerides error on the extracted ranging measurements. The adopted selection metric considers the angle between the estimated receiver-to-SV range vector and the SV’s velocity direction. The paper demonstrates high correlation between the derived metric and resulting ranging error. Monte Carlo simulation results are presented with pseudorange measurements from 130 LEO SVs (101 Starlink, 22 OneWeb, 2 Iridium, 2 Orbcomm, and 3 Globalstar) to estimate the position of an unknown receiver via an extended Kalman filter (EKF). Starting from an initial estimate 14.14 km away, it is shown that using all SVs with poorly known ephemerides results in a three-dimensional (3-D) positioning error of 4.19 km, while the proposed selection approach reduces the error to 22.90 m, adaptively using 1–7 SVs at each measurement epoch.