Part of the answer can be found in the thread on current KAL pilot culture. It was certainly no better in 1983, meaning if the Capt. left a switch in wrong position, nobody would dare question it. That said, what are you chances of coming even close to Sakhalin if you had remained on a vectored heading out of ANC?
The LTN-72 aboard that 747 required keying in the departure lat/long and 9 more waypoints. You weren't supposed to do it, but crossloading departure waypoint data from the first to the other two INS was possible.
The easiest mistake would be to accidentally key in 139 or 140 W. Long. instead of 149W for ANC alignment. INS will detect erroneous latitude, due to earth rate, but not erroneous longitude. A second pilot was supposed to verify the data entered, but.. Once the erroneous departure point was inserted, all the following navigation would appear normal; you would just be west of your intended track.
INS based position reporting would have been late, especially while still over Alaska, and their reports were about a half hour late, from what I heard at the time.
Just a month before KAL 007, a DC-10 departed LAX, headed for HNL. They were north of course upon leaving the coastline, and were alerted by ATC. They returned to LAX and wrote up both INS. That vintage INS had non-volatile memory only for the last position when shut down, which was also departing point in this case. In other words, with operable INS, the last position displayed was what had ben entered before departure.
The memory in both units showed Lat of 33-56N, which was correct, but Long of 104-53W, instead of the correct 123-3W. Where did they get 104-53? Looking on a map, that position was in Oklahoma! It took only a moment to look straight north and see Denver. DEN was one of their hubs. A pilot had entered the wrong Long from his faulty memory, and the other crew didn't catch it.
Had they continued on, great circle nav would have taken them well north of Hawaii, with a destination 800 miles to the west.
GB