If this does end up going to a super-synchronous transfer orbit, then it will provide an interesting comparison with the Echostar 23 launch. Echostar 23 was ~5500kg and was sent to 179km x 35903km x 22.43o, which works out to about GTO-1711. Inmarsat is currently listed as 6100kg. If they have enough performance in the F9 to get it to super-synch, this may be the first definite upgraded performance launch (there seems to be a bit of debate whether NROL-76 used upgraded thrust or not).
This might also depend on the desires of the customer. If you have more performance than you need to reach a minimal GTO (GTO - 1800 for the cape, 179 x 36000 x 28
o) , you can use it in two different ways.
(a) Keep the apogee at GTO, but reduce the inclination. Here they reduced it from 28
o to 22.43, saving the customer about 100 m/s to get to GEO.
(b) Increase the apogee, but keep the inclination. With the same delta-V, they could have generated an orbit that was about 179 x 48000 x 28
o (super-sync). In the case of Echostar this would have been a very similar reduction in delta-V.
Option (b) involves more navigation (at least two burns and longer intermediate orbits) by the customer, and the satellite has to be capable of working when above GEO. But in some cases it can be significantly less delta-V required by the launcher. In general, (a) is better if you have only a little extra performance to use. As performance increases, the two techniques become roughly tied at about GEO-1700 from the cape. But for better orbits than that, the super-sync method (b) pulls ahead. For example, the same delta-V can give you 179 x 119,000 x 28 (for GTO-1500) or 179 x 36000 x 17.7 (for GTO-1625).
From the point of view of SpaceX, the only difference is the pointing of the second stage at the GTO injection burn. It's at the same place and lasts the same time in either case. So I suspect they offer the option to the customer, and the customer chooses based on the navigation required, the delta-V saved, and the capabilities of the satellite.
Here is a plot of the effectiveness of both approaches, starting from a 180x180x28
o parking orbit.

EDIT: Point out that (a) is better for small changes, (b) better for large ones. EDIT again to add plot