One is 0.37a and another one is 0.32a, which one do you rather choose to run your 18w sox?

And what the hell, the 20w ballast with 0.32a is as big as the 36/40 watter

They are both basically the same ballasts, one is more efficient then the other, now if your going to run a SOX on them measure the current going to the bulb and keep it with in range or the lamp will suffer because running then either to high or two low will affect there life span, and you may not get a high enough open circuit voltage to fire the tube up from a simple reactor ballast (It might work with no problem but it might not they do like about 300 volts to start reliability, ideally more like 425, one trick to get them to fire is simply turn it on and off until the lamp starts up which is ok for testing but not for dusk to dawn use and you might need to add an igniter or starter if that is the planned application) and if your in the UK look at the used market it's not that hard to get the correct ballast for a SOX over there which would be ideal...

A parallel SOX electronic ignitor will do the job of delivering sufficient voltage for starting.
The upper ballast seems to be closerin the current (18W SOX are rated 0.36A, if I remember well), but the 0.32A is for sure not that far away for the life to suffer (just the efficacy/output may get a bit lower, but not that much), but it will be cooler running.
So both should be good enough.

The 18W ballast can also run 26W PL-C and 28W 2D, and that indeed runs at 0.32A on the same ballast. It is possible that the etch was edited from PL-C or 2D to T8 (without any actual change to the ballast) but they forgot to update the current rating as well. Connect an 18W T8 and measure...

The ballasts are the same in regards to losses, as can be seen on them (which is unusual in its own right, the losses aren't normally stated on the ballast). The bigger ballast uses thicker laminations (probably as result of using crude production equipment that can't take the thinner laminations), as such, they have higher eddy current losses. To bring the losses down, the core is made bigger in order to lower the flux density