Can't tell it the mainspring is totally wound or not at all .
How can I tell?
When you push the backspring in and it does not come back out, then the mainspring is wound as far as it will wind. And if everything else in the power train is good, the mainspring will be fully wound. And ideally, you should not push the backspring any further than the bellows would push it, determined by the depth of the bellows when the room is at its warmest temperature. I cover the spring guide (disc) at the rear of the backspring with the cap from a can of spray paint to keep from pushing too far.
Keep in mind that when the backspring does not come back out, the bellows will no longer have any impact until the mainspring lets down some. PULLING the backspring to force additional winding does NOT duplicate the action of the bellows, and leads to bad data on how well the clock is running.
Also keep in mind that the mainspring of an Atmos does not get wound all the way up like in a wall clock. It only winds to the point of equilibrium with the action of the bellows as delivered through the backspring. The intent is to maintain consistent power delivery. Since the mainspring is wound only on the collapse of the bellows, equilibrium is largely determined by the strength of the backspring. This points out the importance of the length of the chain and where it is pinned on the spring guide. The distance from the spring guide to the rear movement frame should be 45-48mm when the mainspring is all the way down. But to make this measurement, the movement must be removed.
Because the chain is twisted, we already know the power train has been tinkered with. I encourage you to avoid making any adjustments to the time train, fork/roller, etc, until the power train is validated. Removal of the movement is simple, but you must hold the intermediate wheel to prevent rapid let down of the mainspring. Reinstallation is also simple and provided the movement, particularly the fork, is treated with care, will make no change in the condition of the clock.