Appears to be an issue with the underlying parser when the whole thing is in hiragana, and there is no other sentence structure to add context. https://ichi.moe/cl/qr/?q=%E3%...
I doubt it's a greedy algorithm, so it most likely (without know the internals of the engine) is that it isn't ever "reaching" that far down the string before it makes the decision that の should be separate.
Maybe I don’t understand how this works, but it seems to me that if I follow the link you provided above, and change the options to (kana; no segmentation), then it ought to do the right thing. The fact that it doesn’t leads to believe that the problem is not the segmentation algorithm, but the dictionary lookup.
