I searched and could not find much help on what the differences between the Roman and x-Latn sections of the name form.
Those fields are actually a result of old data/formats, before the current language templates were added, and I think this is a bug. Something in the new page's coding is doing Weird Things with the data from whatever automated process was run in 2016.
There's a group for discussions and bug reports on the new page (https://community.familysearch.org/en/group/316-new-person-page). I found your example (https://www.familysearch.org/tree/person/details/KW6B-VYM), shall I report this, or would you like to?
I've been correcting all of these as I run across them out of fear than any automated process for doing so will just make things worse even though it does not appear that the format causes any problems. At the same time, I set the correct language flag if available instead of having Other be there. It only takes about 15 seconds of copying and pasting to do so
I will report it.