On natural attacks
Despite seeming absurdity, the spell undead anatomy I et al. grants the natural attacks that the spell says it grants, and that means, no matter the form, a creature that's the subject of the spell gains the bite attack and either two claw attacks or two slam attacks. Is that weird and a gross simplification? Sure. But Pathfinder mostly succeeds in making reasonable many of the form-changing issues of its predecessor.
(That predecessor—Dungeons and Dragons 3.5—has form-changing rules that underwent so many changes that only the game's most dedicated scholars can determine how such effects work. Some folks—like me!—, rather than deal with the rafts of errata and frequent outright rewrites, just ban form-changing altogether in D&D 3.5 except in specific cases. But in Pathfinder there's no need for such a heavy hand.)
So while it is weird and a gross simplification and maybe sometimes dumb, the effect is also playable. Just getting those specific natural attacks—even when such attacks probably shouldn't be gained—from the spell undead anatomy I et al. instead of, for example, whatever natural attacks the creature actually has (or, anatomically, should have) means Pathfinder's monster designers needn't limit their monster designs. That is, if Paizo wanted to publish, for example, a 2 Hit Dice undead with 10 natural tentacle attacks, Paizo can and needn't worry some PC is going to change into that crazy form by casting the spell undead anatomy I et al. and get all those tentacle attacks.
On the assumed form
The spell undead anatomy I et al. is specific about transforming into undead and that includes in the spell's examples templates that transform a creature into a creature with the the type undead, like the template skeleton and zombie, contrary to the rules of the school of transmutation. Paizo threads from 2012, 2014, and 2015 lament the same lack of clarity on precisely what forms can be assumed by the spell undead anatomy I et al. that your question does. Currently, how deep the rabbit hole of what forms the spells allow one to assume goes is a GM's call. Does the spell allow a human to assume the form of a goblin bodak, a monkey skeleton, an oread lich, or a troll zombie, all of which are, as the spell requires, "vaguely humanoid shaped"? Maybe, and, while there doesn't seem to be anything particularly unbalancing or wrong with a GM ruling that those forms're okay, exactly what forms can be assumed is still a GM's ruling.
However, were I GM and wanted to make things easy on myself, I might make a house rule that the spell undead anatomy I et al., while allowing changing form into a form with a template, only allows changing form into a creature with only one template and that template must change a creature's type to undead.