In my experience, small parts are not more or less accurate. That is because the Maslow has shown excellent repeatability and is thus very precise.
There are several sources of errors that you can control:
a) frame and sled stiffness
b) warping of the workspace area due to 12 O’clock and other machine details
c) router bit retroaction forces
You might want to look at the zipper three challenge and follow thread links to see how to improve your Malsow setup.