The use of ateji does not always disregard the semantic content of the characters.
It's true that many ateji use the phonetic content of individual characters to represent words, ignoring the meaning of those characters. But sometimes the characters are chosen so that the meaning is also relevant. This is known as phono-semantic matching. Wikipedia gives the well-known examples of () and (), where both the semantic and phonetic content are matched. I haven't done a detailed search of the etymology of , but if the semantic content of the characters is related to the overall meaning of the word, this does not preclude it from being assigned as ateji.