A one microsecond variation within the airborne equipment's processing circuits could easily result in a 2,000 km error in the position calculation.
Light (and radio waves) moves quickly. I cannot keep up with it.
But. The speed of light is approx 3 x 10^8 metres per second. One microsecond is 1 x 10^-6 seconds.
So, in one microsecond light (and radio) will travel approx 3 x 10^2 metres.
At an "altitude" of 40 degrees - lets change that to 45 degrees to keep it simpe - a 300 metre variation in the hypotenuse would mean, roughly, a 200 metre variation in the horizontal position.
Is my maths wrong? If not, hardly 2,000Km.
And as has already been posted, while the equipment may not be bench calibrated , Inmarsat are very likely to have had similar "ping" data from when the plane's postion was known with considerable precision - as when on the ground - to calibrate their analysis.
There will, of course, be variances between different pieces of equipment. An individual piece of equipment will also have variations over time. But do not underestimate the sophistication of the equipment, particularly of the complete end to end comms system including the plane mounted equipment, the orbiter and the ground station. Or the capabilities, experience, and intelligence of the people that have carried out the analysis.