It's difficult to tell without slowmotion, but I think there are probably some sort of waves and resonance in the chain's motion. Simply by that, with enough slack, the drivers could repeatedly jump out of the groove (at least, partially) and go back in. A slight lateral movement of the bar or wood, and the drivers can't found the groove again and slide out of the bar (or is more easily pushed out sideway).
The chain is straightened between the wood and the sprocket, as the sprocket pulls the chain and the wood keeps holding it.
With wood on the lower side, the straightened portion of the chain is short, say 1/4 of the total chain's length. The slack is spread on the 3/4 length between the sprocket toward the wood. So there is less slack per driver, aka less possibility for this one to jump out of the groove.
With wood on the upper side, the straightened portion of the chain is long, like the 3/4 chain's length, going around the bar nose. In this case, the slack is confined only on 1/4 of the length, each driver in this area has more liberty and is more prone to wobble / jump out of the groove.
It's only a thought, I don't have evidence of that.