Proximity operations around Phobos are critical for the Japanese Martian Moons eXploration (MMX) mission, to be launched in 2026. Accurate knowledge of the spacecraft's state along its nominal quasi-satellite orbits is required to perform scientific observations and better characterize the dynamical and geophysical environments of the Martian moon. This paper presents a consider covariance analysis implemented in the ESA flight dynamics and mission analysis software GODOT and MIDAS to quantify the knowledge error of MMX in its lowest altitude trajectories. Different observables are considered, including radiometric tracking and onboard lidar and camera measurements. The latter are simulated using an improved synthetic image generation pipeline enabled by the Celestial Object Rendering TOol (CORTO), which was optimized based on real Phobos images. The covariance analysis shows that the relative knowledge of MMX with respect to the Martian moon can be known with high accuracy (less than 100 m, 3-σ) when estimated with all of the observable data.