The article you are quoting is asking for a test of Charge conjugation, Parity transformation and Time reversal (CPT) invariance at high energies at the LHC and presumes that violations of this will be the signal for the sought for black holes of large extra dimensions. This violation is argued on the lines I am arguing against, below, and certainly is not a strict prediction of such large extra dimension models.
I am a simple minded experimentalist having worked with the analysis of various experiments in the large CERN collaborations for forty years+. The brain gets cluttered with all the details of detectors and their errors, programs and their errors so when it comes to reading theories
a) I keep it simple
b) I try to think of an experiment that would validate the theory.
As far as elementary particles and General Relativity goes, a) tells me that General Relativity says nothing about elementary particles, and a subset of GR, special relativity, applies to elementary particles and the theories evolved to describe and predict the data.
Now CPT invariance of a function is a theorem that applies to the theories and the functional forms developed to describe elementary particle dynamics, very successful theories.
Elementary particles come one by one to start with. The collective behavior of systems of elementary particles has been studied and functional forms found, but still the theories DO NOT contain gravitational effects. CPT has been tested and holds for the experiments we have done.
Is the thought experiment of a neutrino crossing an event horizon a valid thought experiment?
To start , in my simple minded experimentalist's view, we are on dangerous ground because we are mixing GR ( event horizon) and elementary particles. That is, at best, we will have a semi classical ( hand waving) approximation of what may really be the case. Can one test the validity of a theorem validated in particle physics while mixing GR and quantum mechanics?
My intuitions says not, if we go by precedent of such semi classical approximations. They are not a strict theoretical tool and are heuristic. They may seem to work as in the Hawking radiation case because no contradictory data has been found but to base a violation argument on the approximation seems to me a game with many probable outcomes.
Suppose when the four forces are unified in a quantized theory of everything ( TOE) the event horizon comes out to be quantized, as an example of a bizarre outcome. It will be a complete game change. In fig2 of the neutrino example, how do we know that the black hole does not increase its lepton number by 1 if we do not have a quantized theory of gravity? After all an electron captured in a nucleus keeps all the quantum number sums correct.
We were at the same impasse with the black body radiation and thermodynamics, the ultraviolet catastrophe. When a complete model of quantum statistical mechanics was developed the catastrophe disappeared. ( actually it was the reason the quantization of photons was proposed). So I am proposing, in this semiclassical handwaving, that the black hole gains a lepton number when a neutrino falls in ( or a baryon or whatever) and that each black hole is characterized ( is in an eigenstate of ) by all the quantum numbers falling in it. I can hand-wave with the best of them, as an experimentalist.
We shall reach b) if the experiments trying hard to find large dimension black holes in the LHC have some candidates. That will be fun, checking the quantum numbers and designing new experiments.