According to the link "Strava uses your power output and a coefficient for human efficiency. " Checking a recent ride of mine, the coefficient used is about 0.25 ie you burn 4* as many calories as you put out power.
I find the Strava calories pretty consistent and reliable. In essence I think they work by using an aerodynamic coefficient, a constant resistance and a velocity dependent resistance. There's a description of the type of model used for this sort of thing here.
https://ridefar.info/wp-content/uploads/positions.jpg
Strava also takes into account the power gained or lost from gravity, so will subtract (or add) the energy gained or lost in climbing/descending.
254kcal in 2.6 miles sounds like a *lot*, and it's on top of the energy gained from descending so I'm guessing there is some glitch associated with that figure. It equates to 100 calories/mile, which is double what I clocked on a recent hilly ride. Have a look at the analysis. 30 kcal sounds realistic for a downhill 2.6 miles.
On a generally downhill ride, you can expect the figures to fluctuate a lot from ride to ride, because Strava is essentially guessing what your resistance is, and if it's close to your freewheeling speed, a small amount either way will make a big difference to net calories.