B757 crews dealing with an AUTO SPD BRK EICAS after takeoff discovered an unacceptable IRS navigation deviation during delayed navigation accuracy check. The flight returned to land overweight.

Narrative

I was the non-flying pilot in May 2009. While climbing out I was completing checklists and speaking to Maintenance and Dispatch about an AUTO SPD BRK EICAS we got on strut extension so I was completing my ETOPS checklist a little later than usual. We were 100 NM from our departure airport when I began the gross navigation error check. The FMC showed a 4.5NM difference from the raw data DME readout; the bearing checked right on; using departure station VOR. My First Officer then checked the same thing on his side and got the same result. We then tried to use our destination VOR as it was closer and got the same results. We then pulled up the IRS POS page on the FMC and went to the bearing/distance prompt and saw the IRU's were showing a 4.2 NM to 5.6 NM discrepancy. This discrepancy was increasing with our distance from shore. After speaking with dispatch and maintenance again we decided to turn back for our departure airport. We made an uneventful overweight landing with equipment standing by.