Hong Kong Orchid Tree:
Plantae / Angiosperms / Eudicots / Rosids / Fabales / Fabaceae / Bauhinia x blakeana
Cherimoya:
Plantae / Angiosperms / Magnoliids / Magnoliales / Annonaceae / Annona cherimola
Their closest common grouping is "angiosperms", aka all true "flowering plants", and excluding cycads, ginkgos, conifers and gnetophytes. Aka, pretty bloody well unrelated. Cherimoya is more closely related to the parasitic carion-scented Corpse Flower than it is to Hong Kong Orchid Tree.
And Hong Kong Orchid Tree is more closely related to cacti than it is to cherimoya!
Leaf shapes are one of the worst ways you can attempt to judge how closely related plants are. Leaf shapes can vary highly between species in the same genus, and can even vary between young and old plants. One of the most extreme examples of plant leaf variation (getting into trivia here!) is Boquila trifoliolata , the Chameleon Vine; its leaves not only dramatically change in shape and colour as it grows, but they do so to
match the leaves of whatever plant it's growing on. If a single vine grows across multiple trees, the leaves it makes on host it climbs on will match that host's leaves.
How it does this is still unknown. It doesn't appear to be "pre-programmed" based on chemical cues; it'll even adapt its leaves when growing on a fake plant. The camouflage ability is so effective that it was only relatively recently discovered; people just assumed that it was a highly variable species. It's edible, by the way, although I've not heard any reports of the fruit taste.
Anyway, back on topic: the genes that determine leaf shape mutate readily. Which really, you'd expect, wouldn't you? If a plant moves into a new environment , the leaves are one of the first things that needs to change due to the different light / moisture / wind / etc conditions.
Lastly: I don't even get your premise. Hong Kong orchid tree leaves are nothing like cherimoya leaves. Not in shape (double lobed vs. single), not in vein structure (palmate vs. pinnate)... they're totally different.