In general, because G' is a signed quantity, I would expect any experiment to be quoting double-sided errorbars. If G increases, one thing happens, if G decreas

In general, because G' is a signed quantity, I would expect any experiment to be quoting double-sided errorbars. If G increases, one thing happens, if G decreases, the opposite happens, so the difference is discernable. I'd take 0.0 to be their reported central value.