A script that was written by hand for six hundred years in two divergent traditions does not have a fixed inventory of shapes, and any list of its letters is a decision about where to draw a line. This sign is on the line.
Three quite different things are being counted whenever a number is given for the Glagolitic alphabet. There is the recited alphabet, the sequence of named letters that a learner memorised, which has a definite length in any one tradition. There is the repertoire of distinct signs attested across the surviving manuscripts, which is larger, includes regional forms and positional variants and grows whenever a new fragment is identified. And there is the encoded repertoire of the Unicode Standard, which has to provide an address for anything anyone might want to reproduce, and which is what the atlas follows.
None of the three is wrong and none of them is the alphabet Constantine devised, because no witness to that survives. The atlas therefore lists forty four signs, marks the variants as variants, records the numeral values only where the tradition assigns them, and states in an open question what evidence would be needed to reconstruct the original inventory. Saying which question a number answers is more useful than choosing a number.