I suspect..
..that it's based on the mainframe view of files. Also encountered in the VMS world. Basically these OS provide the means to identify the content/structure of a file. VMS for instance has attributes that can indicate things like "CR/LF text", "LF text", "Bucket database"(*). This isn't just like Windows' extension recognition. You actually can't load a text file into a text editor under VMS unless you change the attribute to indicate the correct file structure.
Under windows "ren *.doc *.xls" is perfectly valid. The equivalent under VMS would be tantamount to corrupting the file.
By that definition everything Windows does is 'unstructured' because as far as the OS is concerned every file is just raw binary data. It's up to the application processing it to know how to read it. If you want to get NotePad to try and open an Excel sheet you just have to force it to by specifying the extension manually. Doubtless you could code a VMS app to do that if you wanted but it would be perverse.
NB:I haven't used VMS myself but I did used to be involved in data recovery and I was responsible for writing the Files-11 extraction tool. That taught me that copying the files from a Windows machine to a VMS machine was not enough. They need to have their attributes fixed up once they get on the VMS machine to be of any use.
(*)Not sure about that one but VMS does include database functionality in the OS.