An aircraft with constant bank is not rolling, period. This is true regardless of weather neutral ailerons are required in the turn, or aileron is required to into or out of the turn.
Which way the airplane will "break" once stalled has nothing to do with the bank. Airplanes may break "over the top" or tuck under by rolling to the inside, depending on several factors.
One of the most decisive factors is rudder useage. Some feel that the rudder deflection is what what breaks the airplane one way or the other, but that's not it.
When yaw is applied outside of a balanced condition in flight (inclinometer ball in the center), a yawing action is introduced. This may occur due to rudder input, or lack of rudder input. This may be due to adverse yaw owing to aileron deflection. Many possibilities exist.
The inclinometer tells you if you're slipping or skidding. In either condition, you will experience not only chordwise airflow over your wing, but also spanwise. The airflow patterns over each wing are altered, and are not symmetrical one wing to the other. Reduced aileron effectiveness will exist on one side of the airplane, and reduced airfoil effectiveness on the other, blanked slightly by the fuselage, with additional interference drag from the fuselage.
Put in left rudder in a left turn, and skid the airplane. You have options. You can apply right aileron, lowering the left aileron, increasing adverse yaw to the left and induced drag on the left...you'll get a left break in most cases. Additionally, the left wing will be faced with increased drag, and reduced airflow over the inner wing and more spanwise flow over the outer wing, increasing the effect of the aileron. The right wing, already experiencing reduced drag from the aileron reflexed upward (to a point) and increased spanwise airflow, will be operating at a reduced aerodynamic angle of attack.
It's important to realize that AoA represents more than the simple angle between the chordline of the wing and the relative wing. Local AoA varies from that of the free airstream, and configuration changes alter the aerodynamic chordline of the wing, alterning the critical angle of attack, or the point at which the stall will occur.
It's also important to note that the wing does not stall evenly, nor does it experience a uniform change in angle of attack. Local AoA is affected by the wing shape, thickness, etc. It is affected by the wing structure, and proximity to other parts of the aircraft. The wing structure adjacent to a nacelle, tank, pod, etc, will experience different airflow than the wing two feet from that item.
Wing planform has a lot to do with the stall behavior to be expected. Large rectangular wings will exhibit different general stall behavior than eliptical wings, tapered wings, or modified taper. The addition of stall modification devices such as stall strips, boundary layer energizers, vortex generators, slots, slats, etc, will all make a difference. The assymetrical effects of changes in these devices makes a great deal of difference.
Finally, no airplane is perfectly symmetrical. Slight variances in structure, past repairs and damage, or the simple aging of the airplane leads to subtle differences one side to the other. Some airplanes, all else being equal, will tend to break one way or the other, where an identical make and model will not.
When I was eighteen, I began flying ag, crop dusting in Cessna AgTrucks, Pawnees, etc. My second operator had three trucks, each of which were identical except for color. We operated these airplanes heavy, meaning close to gross, with minimal performance. We made low level steep turns in them, such that feeling the stall buffet in each turn wasn't uncommon. Because of the proximity of the ground (75') in the turn, a strong association was able to be made between what the airplane was doing, and the slipping and skidding action. The perephrial image, as well as the buffet and the other things associated with the turn, lead to a very sharp intuitive sense of what the airplane was doing, and needed to make it do something. In short, seat of the pants type flying.
I found out very early on that each airplane had a personality. They did not fly the same. For someone flying them around the pattern or taking them on a cross country, the difference might not have been noticable. However, we used the full operational envelope for the airplane on every flight, often every 30-60 seconds during the course of a flight, and the differences were painfully obvious. I've noticed this time and time again when flying for companies or operators who have a fleet of similiar of seemingly identical airplanes. Each one has distinct personality.
Simply because an airplane appears to break left or right, does not mean that a left turn causes an airplane to break over the top. It's not true. Most likely I think the origional poster needs to look at aircraft control to determine the cause. I believe that upon inspection, he or she will find that it's operator error causing the break. Secondly, he or she needs to compare this behavior to other similiar airplanes. If no difference is found, it's still not evidence that an airplane breaks over the top in a left turn...because that's not true. It's then evidence of operator error...the real reason that the majority of airplanes that do break one way or the other, do so. Have a ball.