"that sounds like rather a large difference to the inmarsat analysis."
As I understand it, all that's been published about Inmarsat is arc position and direction. Anywhere south of the last known radar position and on the arc for the last handshake, (NOT the time the plane's engines went off, which might be up to an hour later) is still consistent, I believe. Also, there will have been drift from wherever the plane did come down, if at sea.