SimpleHelpers.Net | Micro-libraries for .Net | Functional Programming library

That's probably because of the way the code you linked (ObjectDiffPatch) builds up the array diff object. This is the relevant part, as I recreated it in LINQPad:

I added this as an image because that tooltip is important here. But let's start at the top:

If one of the arrays is empty, it properly keeps the original structure: grabs the two arrays (one of which is empty), and adds them under the field name "Companies".

If, however, neither are empty, then it starts comparing their contents, piece by piece. And when a difference is found, the code adds it to the diff object using AddNewValuesToken, which expects a field name as its third parameter, but here we don't have a field name, just an array index.

Now, the code "solves" this by converting the array index to string, and treating it as a field name. No wonder that the resulting JObject will treat it that way too – this is why you see an object in the output, where what used to be array indices, become fields.