Impact assessment methods calculate the midpoint or endpoint impact from the release of environmental flows. If you only care about carbon dioxide, then you don't need an impact method. If you want to calculate the midpoint impacts, i.e. global warming potential, then you should use a method that's consistent with your data. If you are using ecoinvent data than you'll need to use that method within the openLCA methods pack. In most cases, the results from impact assessment for global warming potential should be pretty consistent across impact methods outside of slight differences in the methodology used (e.g. IPCC AR5 with or without feedback vs IPCC AR4). The "best" in this case is one that uses the environmental flows with which you are modeling and best matches intended goals.