I'm totally new to this part, just started digging into it this month. I'd be very presumptuous to say there is a problem with the options generation...
I'm probably not able to verify arguments -L/-H computed from the telescope and ccd variables because I don't know the maths. But I know the scale boundaries in arcsec/pixel of the ccd+telescope setup from successfully solving a picture on nova.astrometry.net. Thus right now I tend to replace the options with that calibration.
The other thing is that nova.astrometry.net will search a scale from 0 to 180 degrees, with a downsample of 2, and that's all, probably because they have a relatively strong host behind. If the telescope/ccd parameters are not filled in correctly in Ekos, or if something on the optical path is not taken into account, the optimization in the arguments will probably drop the actual solution.
And finally, nova.astrometry.net will search in the index list called the 200-series (see for instance
groups.google.com/forum/#!topic/astrometry/RfpNop2fTV8). The offline and remote solvers will use what is called the 4200-series. Right now I don't recall which star catalog gives which series.
I have no numerical example at hand to go with this post unfortunately. I'll work out one as soon as possible.
-Eric