Diversification;
I assume than one or more of these were in error due to the errors in the speed measurements, which may have grown slowly over time. In my personal opinion, this may well explain why the transmitted FM-position perhaps was more and more off from the correct position near the end of the flight.
While there are always errors slowly introduced in the IRS calculations, the flight was only into its 3rd hour. Unless an IRS began failing, IRS drift and therefore the calculated bias to produce mixed position would be quite small. The most drift have ever seen in years of overseas flying on the A330/A340 was in the neighbourhood of 20nm. We can see by the chart that a nominal drift at 3 to 4hrs into the flight could possibly be around 7 to 12nm or so and that would be significant. In reality the drift is usually far less. The IRS position bias is updated by radio or GPS but the original IRS calculated positions are not so updated. I am assuming that FMS position errors would always be less than IRS errors.
PJ2