So you reckon they are simply referring to different points on their radiation patterns?
One being optimistic and one being conservative?
Still seems like a huge difference in range, wouldn't you think they would both use their maximum value?
More likely their definition of "range" is just different. Does it mean "distance in free space at which communication is still possible with a < 50% error rate" or does it mean "distance in 'real life' conditions at which communication at full speed with low error rate is likely"? I'd guess one or the other has a looser specification on what the expected data and error rates will be at full "range".
Anyway, you're kind of comparing apples to oranges here, since the XBee you're referring to is a ZigBee radio and the Nordic module is not. ZigBee is only 250kbps while the Nordic modules do 1mbit or 2mbit, this will add a fair bit of noise margin. The modems will also likely be implemented differently, and possibly the radios themselves have better (lower noise / more sensitive) receivers. This is clear from the specsheet sensitivity numbers - Nordic claims -85dBm at 1mbps, Digi claims -102dBm.
Really these numbers are marketing wank though, not an engineering specification, they can practically just make them up. There are too many, and too loosely-defined variables to glean anything other than order-of-magnitude information out of them. What you need to do is measure (or look in the specs, if provided) and find out what the minimum signal strength at the receiver is that results in acceptable performance for your application. Then you can estimate your antenna gains & path loss and figure out what an approximate range for your usage is going to be.