Payt writes an ejective p, and there is no Semitic consonant behind it. Proto-Semitic had a plain p, which Ge'ez turned into f, and it had no p of any kind at the ejective series where b, d, g, z, k and t all have partners. The labial column of the inventory is the one place where the pattern that organises the rest of the fidel simply stops.
So the presence of two p letters at the end of the Ge'ez order, payt and psa, is something the inheritance cannot account for. Both stand late in the sequence, which is where added letters usually stand, and the sound they write is marginal in the classical language, occurring chiefly in names and words taken from Greek.
Where they came from is not settled. Payt has been explained as a sign made by modifying tayt, on the resemblance of the two shapes, and psa has been explained as a borrowing of the Greek pi. Neither explanation rests on a document; both are readings of the letterforms and of the order. The atlas records the relation between payt and tayt as debated for exactly that reason, and carries an open question naming what evidence would settle it. Recording the argument is worth more here than choosing a side of it.
The atlas therefore records twenty-six letters and describes four of them as additions: two inherited from a South Arabian inventory richer than the Northwest Semitic one, and two supplied later for a sound the language had lost. That is a more informative statement than the count on its own, and it is available only because the order preserves the sequence in which the letters arrived.