While I am not on an US-style admission committee, which I suspect this question implies, I assume that the answer is quite simply that the committee does not have enough time. Or, in other words, the information tensity of taking half a day off to read a single thesis with any thoroughness to better evaluate a single candidate is not high enough.
For better or for worse, any evaluation on any level (PhD admission, faculty applications, paper reviews, grant reviews, tenure, ...) needs to balance thoroughness with time effectiveness. Pragmatically speaking, this is also why we have (for better or for worse) grades, bibliometrics, awards, honours, conference rankings, or journal impact factors - not that the information captured by these metrics is great, but because they give away some information in no time at all.
To what extent time becomes a factor obviously also depends on how many candidates there are to be evaluated. A tenure committee can and should afford great thoroughness because it is evaluating only one (or a few) candidates. PhD admission, on the other hand, usually needs to compare hundreds of candidates. Note that, annoyingly, the importance of the decision to the candidate is usually not a substantial factor in the decision how much time to spend.
That being said, I personally still often use the master's thesis as an important piece in my decision for or against a PhD candidate (I work in Europe, and we hire students directly). However, I am not reading the thesis end-to-end, but I glance over it to find out:
- How well the candidate writes, especially how well the candidate writes in English.
- Whether the thesis would in terms of contribution hold up in my own university - if this is not the case and I do not know the university that the student graduated from, I will assume that the level is weak. If the contribution is small but I know the university to be good, I will often assume that either the student or the advisor is weak.
- Whether the student knows good scientific practice, in terms of plagiarism, overselling of results, truthful reporting of threats, ...
- If the thesis is in my field, I will also check if the student has a reasonable grasp of the state of the art by looking at the background and related work section.
All of this can be done in very short time - it is not necessary to read more than two or three paragraphs to figure out if the student is able to write English documents, for instance.