Letters of recommendation -- from postdoc or professor? Both?

I am an undergraduate studying pure mathematics in an American university with a top-10 graduate program. I am applying to PhD programs and graduate fellowships this year, and I have a dilemma about who to ask for letters. Maybe this question is too specific for academia stackexchange, but actually I think it is abstractly a somewhat common issue. I have four potential letter writers. Two are very solid (very famous professors who know me well from reading courses / research) as far as I know, and I will ask them. Call the other two potential letter-writers "Prof." and "Postdoc" "Postdoc" is a postdoc whose graduate-level course I TAed; I also took a graduate topics course with her and produced something semi-original (a new proof of an already-known result) for the final project. She is originally from Europe and did her PhD in the United States at a top-10 graduate program. "Professor" is a very famous professor, though not in the same field as the other three (who are in the field that I will apply to graduate school in). I took an undergraduate course and a graduate course with her; I did very well in both courses and produced a very thorough (and long if that counts for anything) expository final project for her graduate course. I also know she is very serious and enthusiastic in general about writing letters for undergraduates, though I do not know if she actually read what I wrote in my expository final project. For graduate programs, I have the option of including 4 letters. Should I include all 4, or just choose one of ? Should I expect one of those letters to dilute my application and make it less strong? In my view, there are reasons why either letter could do that: for the postdoc, either because she is a postdoc (so some people might completely disregard her letter), or because she is very young and European (so might not be very good at writing letters yet, especially in the American style). For the professor, there is the potential downside that she is not in my field (though could it be a strength to have some diversity?), and that I didn't do anything even semi-original in her graduate course (though my understanding is that for pure mathematics this can be OK?). For some graduate fellowships (e.g. the NSF GRFP), it seems like I can only submit three letters, so I must choose one of the postdoc or the professor to exclude (regardless of whether I ask both for PhD programs). Given this information, is there a clear choice of which one to exclude? Apologies again if the question is too specific. However, I think there are some general themes in letter-asking embedded in my question: (1) asking postdocs vs. professors, (2) asking professors who I haven't done anything original for, (3) "dilution" of letters and whether to ask for 3 or 4 of them, (4) asking a professor whose research area in pure mathematics is different from the one I am applying for

(1) postdocs < professors, but the former can speak of your teaching ability, whereas the latter can only speak to your ability as a student, I'd be tempted to use two references. (3) Ask just two? (4) Not so relevant, but they can't really help, since you merely took their class.

@user2768 I have made the relevant change. I'm sorry (to you or whoever downvoted) if my question is inappropriate or of too low quality for stackexchange. I have noticed other questions of the same type on the site which seem to be well-received, though. What do you mean by "ask just two?" There is a lower bound of 3 letters for most things, and doesn't this contradict what you said in (1) about asking both? When you say "they can't really help, since you merely took their class," do you mean this in general? Even in pure mathematics, and for a graduate course?

I think that the full extent of my question has not been answered on this site: it asks about how to make choices about how many letters to send when you have the option of sending more than the minimum (usually the minimum is 3), and about how to make choices between different people for a 3rd letter, considering factors like whether they are a postdoc or full professor, but also based on whether they are in your field or not, and whether you did anything semi-original for them. Is the answer that the difference between postdoc and prof dwarves all the other factors?

A point of clarity for the GRFP, you can submit 5 letters, but 4 and 5 are only read if one of the first three does not get turned in. Does GRFP fund pure mathematicians?

@AzorAhai--hehim, Yes, the GRFP funds pure mathematics graduate students; in fact that is the main context in which I need to choose which one to exclude (which can just mean moving one to 4th priority).