The sampled voice would be stored in the expansion memory, not the user memory. At least, this is how I think it works; I have no experience at all with this, because I've never owned a model that can load expansion voices.
I think perhaps the earlier models in the Tyros line allowed you to store individual sampled voices in the user memory, and you could create single voice files that included the sound sample along with the voice settings. In fact, I think you could even use the keyboard to create the sampled voice, but I'm not certain.
But as I understand it, later Tyros models, and the other arranger models that have the ability to load expansion voices, require you to create a new voice using the Yamaha Expansion Manager, and you must incorporate your new voice in an expansion pack so it can be loaded into the expansion memory.
As I said, I have no experience with this, so I'm going by what I've heard other people say. I do know that a lot of Tyros owners were unhappy when they upgraded from an earlier Tyros model to a newer Tyros model and were no longer able to create their own sampled voices on the Tyros itself.
But as with anything else, once you know how a system works and what its ins and outs are, you get used to it and can become proficient in using it and in discovering ways to work around its shortcomings.