I know that what I wrote is not the only solution to your problem, but it may be part of it. Things happen so fast with small models that unless you know what to look for, it can be difficult to determine the cause(s) of a given departure from course. Gyroscopic precession can be very noticeable, especially at low airspeed and high pitch rates. Right thrust does nothing significant to correct for gyroscopic precession, since GP only occurs while the airplane is rotating about the pitch and/or yaw axes (even that statement is oversimplified but we'll call it close enough). Right thrust has the greatest effect against spiral slipstream and P-Factor. The ONLY way to correct for GP is with with timely rudder application, or some form of yaw damper (like a gyro in the yaw axis). Or, as is the practice with pattern, make the yaw stability so great that GP is small in comparison, making it relatively unnoticeable. You probably already know this, but many of those full-scale WWI bipes were HARD to takeoff and land, and many, many of them met their end in a ground loop on takeoff or landing. They required very careful and specific techniques to safely and consistently take off and land. Any departure from those techniques was likely to cause a ground loop or runway excursion.

The rest of what you wrote makes sense. I suspect a larger rudder and/or fin will be the best solution, but not being able to inspect the airplane puts us would-be problem solvers at a disadvantage.

Please let us know how it goes, and good luck.