The trig formula and the relationship between RRPM and G-load both apply to steady-state, level turns. Obviously, because of both the limited available extra power and compressibility, our ability to do level, steady-state steeply banked turns is very modest. No way any gyro can execute an 80-degree banked turn without slowing down, losing latitude, or (almost certainly) both. The formulas don't apply in the case of sinking turns, since the rotor then isn't generating enough lift (not thrust -- lift in the sense of thrust-that-opposes-gravity) to hold the machine up anymore.
So the real-world, steady G's that we can produce aren't that high, even in a steep bank -- BUT --
The designer of a gyro airframe must consider momentary loads as well as steady ones. A sharp, sudden updraft, for example, does not increase RRPM in the short run. Instead, it increases the AOA of the blades in the short run. This can produce short-duration G increases larger than the ones we can obtain in the long run. Lift is directly proportional to AOA.
Jim Vanek has cited results from a recording G-meter in the neighborhood of 4 G. These instruments capture events of very short duration. Jim's findings tend to justify the FAA's strength requirements for utility aircraft (I think theirs is +4.4G).
I'd design around a load limit of +4 - 4.5 G, with a safety factor of at least 2. If you crunch numbers on the primary structure of a Bensen, you'll find that it meets or exceeds this standard.
The interesting game is how to allow for fatigue. For example, 6061-T6 goes from an ultimate tensile strength of something like 40,000 psi for a one-time event down to 13,000 or less for a few million cycles. Cycles add up quickly in rotorcraft.