I've found this with GaBi and OpenLCA. To understand, I went through the top 95% of controbutors to each impact. If you do this, you'll probably find mistakes in both software packages.
Also, it depends on which impact category you are using for your carbon footprint, IPCC 2007 (AR4) and IPCC 2013 (AR5) can give results approx 10% different for more methane intensive processes, for example. Also, IPCC 2013 impacts were implemented differently across software packages within LCA, which leads to minor differences in the results.