The Guanyin case exposes a problem that runs through this whole table: deciding when two figures are one.
Buddhism's ordinary method on encountering a local deity was not denial. Hariti, a yakshini who ate children, is converted by the Buddha and becomes a protector of childbirth, fed by the monastic community in perpetuity. Mahakala takes the name and much of the form of Shiva, recast as a worldly deity subdued and bound to oath. Hayagriva is a horse-headed Vishnu given a different character and a different job. Vaishravana is Kubera. Benzaiten is Saraswati with a sword and a snake. In each case the local being is not destroyed; it is taken into the order and given work.
That is worth stating carefully, because it is a Buddhist claim about someone else's god, made by Buddhist authors, and it is not conceded by the traditions on the other side. Shaivas do not accept that Shiva was subdued by a buddha. The atlas records who made the identification and when, and does not adopt it.
Guanyin is the same problem in a harder form, because the absorption ran the other way and inward. Chinese religion did not keep Buddhism, Daoism and popular cult in separate compartments, and Guanyin is worshipped by people who would not call themselves Buddhists, in temples that also house Daoist deities, under a mythology, Miaoshan, that has no Indian source. She acquires functions Avalokitesvara never had, above all the granting of children, in the Songzi Guanyin, the child-giving form. It would be defensible to argue that Guanyin is a Chinese goddess who took an Indian name.
The atlas has not taken that view, and keeps one entry, because the textual and iconographic transmission is unbroken and the tradition itself asserts the identity. But the decision is a judgment call, it could have gone the other way, and splitting the figure into three unrelated entries, which is what the atlas did before this entry existed, was the one clearly wrong answer.