Our world is a quantum world so everything, including any stage of the Hawking evaporation, may only be predicted probabilistically. At the end of its life, a tiny black hole has a mass comparable to the Planck mass and such a tiny black hole becomes qualitatively the same thing as just another heavy unstable species of an elementary particle.
A question is whether or not you know the precise microstate of such a black hole. If you know it, you may predict the probabilities of different final products accurately from a well-defined compactification of string/M-theory you consider (without string/M-theory, you will clearly be able to make no precise predictions of the quantum gravity phenomena, and this is a textbook example of one). If you don't know the exact microstate is, it is still true that roughly speaking, the small black hole emits a thermal radiation.
However, at the very end of the life of the black hole, its temperature goes up a lot. At the very end, the temperature is close to the Planck temperature (the highest possible temperature that may marginally be talked about in physics) so the decay products may include (with a high probability) very heavy particle species, too. Right before it disappears, a black hole may surely produce a pair of top quarks or even heavier particles. There's still a nonzero probability that it will decay to two photons or anything else that doesn't violate conservation laws.
Actually, the probability is nonzero that a black hole emits another, smaller black hole. It's just very unlikely: such a process is essentially suppressed by $\exp(-S)$ where $S$ is the entropy of the emitted black hole.
For macroscopic black holes, such a factor is zero for all practical purposes. However, if you want to emit black holes that are slightly larger than the minimal (Planckian) black hole, the factor isn't hopelessly tiny and the emission of a small black hole is a possibility. Again, any black hole microstate may always be interpreted as yet another species of an elementary particle. Large black holes have a high entropy and the description in terms of "exponentially many new particle species" becomes contrived. However, for the smallest (Planckian) black holes, the description in terms of new particle species becomes a condition for any accurate description of the black hole's behavior.