Guptasana, siddhasana and muktasana are close relatives. All three are low seats with one heel drawn in toward the perineum, and the differences between them are matters of a few inches and of which foot is placed above. The Gheranda Samhita treats them as three distinct postures and the atlas follows the text, but it is worth noticing what that decision implies about how the corpus counts.
The hatha manuals inherit a figure of eighty four postures taught by Shiva, and the Gheranda Samhita opens its own list by reporting that there are as many postures as there are species of living creature, that eighty four are the best of them, and that thirty two are useful in this world. Numbers of that kind are traditional rather than empirical. What actually fills a list is a compiler's decision about how fine a distinction deserves its own name.
Counting closely related seats separately inflates a list in a way a modern taxonomy would resist, and it also preserves information. Had the compiler grouped these three as variants, the exact placements would have been lost, and it is precisely those placements that the seats exist for. Mallinson and Singleton's Roots of Yoga presents the primary descriptions in translation, and reading them in sequence makes the compiler's logic visible: the list is not a catalogue of shapes but a record of distinctions someone thought worth transmitting.