Sorry, no expert here but I've worked with a lot of DIY standards until I landed a few real kits, and am somewhat familiar with the cal constants. And you may be correct in that their data is just that, data and only in their own structure. Kit labels for the soft keys would have to be elsewhere in resources. But since today's experiment, I found that the 2 missing files did not fix the problem. Then I found the limited .csv files. Not all kits are present.
But the problem remains that only a handful of kits from the menus are present in .csv form. The files are named for the type such as 85032F.csv. These kits do not contain the usual data found in kits, delays, c0, c1, c2, c3 and lo, l1, l2, l3. They are frequency and a pair of data, probably in real, imaginary form. This example file is a 85032F kit for a type N connector, not sure of the sex. My real kit is 85033E for 3.5mm (similar to SMA). It has no .csv file and there's no reason for that.
Quick summary:
1. performing a cal with the internal 85033E kit results in way off responses, perhaps because there is NO KIT present
2. entering the proper 85033E constants into the USER1 cal kit results in near perfect cal.
3. If the.csv files are indeed Siglent's implementation, then:
a) many are missing and
b) this is not a good practice
They could/should have been in .xml or delimited text or some other ascii readable format and contained data from Agilent.
Since posting I also repeated this with an Anritsu Tee shaped cal kit I have. I entered it as USER2 and it behaved near perfectly.
You can tell if the cal matches the kit if the arc produced for a short or open can be brought to a dot by entering a port extension equal in time to the delay constant from the kit for that type and sex of connector.
I think what's left is for someone else to repeat this on a real SVA. Then it can be brought up with Siglent.