Ted:

I believe you should verify any excessive upthrust of your engine, too big AOA of the wing, or a combination, which forces the model to have the CG located too far forward.

Those conditions may be fighting each other, inclining the balance in favor of one type of loop.

It is possible that your model requires a permanent trim up of the elevator now.

Try the loop with minimum power and radius; at sufficient height, of course.

Also, try what Matthews just suggested above, in addition to start the loop from going up, straight or at an angle.

Another factor to verify is the actual deflection of the elevator under load (stability of the pushroad under compression).

Also consider the centrifugal effect during the loop, as a force resisting the circular movement, which is applied to the CG and is related to the square of the speed of the model.

That centrifugal force will be stronger for loops at high speed and small radius, being able to reach around 7 times the weight of the model (7 Gs).

It seems that your elevator does not have enough authority to overcome the nose up pitch induced by that force.

Right after you apply down elevator, its camber increases and also the lift produced by the horizontal tail, the wing is forced to reduce its AOA and its lift.

As the wing looses lift and the tail gains lift, the combined center of pressure (center of lift) moves toward the tail, initiating a rotation around the CG (nose down torque).

At this point, the angle at which the downwash or wake leaving the wing hits the tail-elevator is also reduced, limiting and reducing the AOA and lifting capacity of the tail.

If elevator up trim was present, the reachable deflection will be less.

As a centrifugal force pushing up appears in the CG, with a value several times higher than the weight pushing down; the nose up torque created by this force overpowers the nose down torque created by the tail, and stops and it may reverse the mentioned rotation around the CG.

The more forward the location of the CG is, the bigger the distance between the center of pressure (lift) and the CG (stability margin); hence, the bigger the nose up torque against the loop.

The opposite happens during an inside loop.

Regards!!