If after lowering the lever, RRPM continues to fall, the rotor has started to stall. Left roll is likely. flare will push the disc further into the stall. Judicious forward cyclic to reduce AOA is your only chance.
Say both blades are close to stall, forward cyclic is put in, the angle of attack on the retreating blade is increased, now why would a retreating blade be less likely to stall than an advancing blade? When it does stall, the advancing blade doesn't, but the result of the aft cyclic, the disk has started to tilt aft, and the retreating blade stalls, which would have the same effect as a full rotorstall because the drag is dramatically increased. You can demonstrate rotorstall close to the ground, it's dramatic, 75% and boom, all RPM is gone. Since you're traveling forward, and the retreating blade has stalled, the rotorsystem will blowback and chop the boom.
Quote from SN-24
"When the rotor stalls, it does not do so symmetrically because any forward airspeed of the helicopter will produce a higher airflow on the advancing blade than on the retreating blade. This causes the retreating blade to stall first, allowing it to dive as it goes aft while the advancing blade is still climbing as it goes forward. The resulting low aft blade and high forward blade become a rapid aft tilting of the rotor disk sometimes referred to as "rotor blow-back". Also, as the helicopter begins to fall, the upward flow of air under the tail surfaces tends to pitch the aircraft nose-down. These two effects, combined with aft cyclic by the pilot attempting to keep the nose from dropping, will frequently allow the rotor blades to blow back and chop off the tailboom as the stalled helicopter falls."
Tim Tucker had the RPM in the certification tests they were doing for the R22 down to 73%. I'd like to ask him, but my guess is he didn't have aft cyclic in when recovering. Too bad they didn't have mini video cams back then, would be some very interesting footage.
James