Clearing the answer register to zero had always been a stepchild of Comptometer design. Felt's earliest patents make scant mention of it and the drawings show only a single clearing knob on the left side with no further detail. The emphasis was understanably on the more complex matters of columnar carry, dial overshoot, etc.
Before the first fifty "woodies" had been delivered however, the design had been refined to include the clearing lever and stop post that would be required in actual production. This redesign would not appear on a patent until 1903 coverng the metal case that would engender the famous infringement suit brot by Felt & Tarrant against Burroughs in 1912.
Until the appearance of the H-model in 1920, the clearing handle had been located toward the rear of the case and required a backward and forward action. The H-model dramaticly improved this rather awkward operation by moving the pivot point of the handle forward and requiring only a flick of the operator's little finger to clear the machine. Since the lever was spring-loaded, it automaticlly returned to its "cocked" position, ready for the next clearing action.
Observing the (rather crude) animation below, I'm convinced that this machine was, indeed an early prototype for that change. The pivot point had not yet been moved forward nor had the mechanism been brot "inboard" the case. Also, the spring-loading was woefully inadequate to the job.
So was this machine the missing G-model? Was it, perhaps like the E-model, an alternate approach, pursued in parallel that lost out to the H-model's elegence?