Almost every Tibetan word an English reader meets in print arrives through a transliteration rather than in Tibetan letters, and the commonest of these systems is the Wylie transliteration, standard in scholarly writing since the late nineteen fifties. It is designed to be reversible: every Tibetan letter has exactly one roman equivalent, so a Wylie string can be turned back into Tibetan writing without loss and without guesswork.
That design has a cost, and kha is where the cost is clearest. The h in kha marks aspiration on the preceding letter. It is not a separate sound, it is not the h of hat, and the sequence is not a fricative. The same convention accounts for cha, tha, pha and tsha, none of which is the digraph an English eye expects: tha is a t released with a breath and has nothing to do with the th of thin, and pha is a p and never an f.
Because Wylie is a transliteration rather than a pronunciation guide, it also declines to tell a reader what a word sounds like. Silent prefixes, silent superscripts and silent suffixes are all written out, so a syllable spelled bsgrubs in Wylie carries six roman letters for a spoken syllable of two or three sounds. Phonetic renderings exist alongside it and are used in journalism and on signage, but they lose the spelling and cannot be reversed. The atlas follows the transliterated forms, because the letters are the subject.