They're not claiming 2+2=3, they're showing that the same methodology that proclaimed 2+2=5 also proclaims (as best as they can determine give the flawed documentation of the methodology) that 3+3=7, 1+1=3 and 17+17=137