Well, I'm writing "scientific" software, so obviously I use lots of +-*/ , often functions from standard math library, and sometimes more advanced math libraries such as LAPACK. Yes, knowing some math is useful.
But: still, 90+ % of the time goes to something else: trying to understand users and their needs, figuring out how to make the user interface and workflow intuitive, working around all sort of bugs and issues that have absolutely nothing to do with scientific algorithms or anything sexy: Why is the firewire adapter randomly corrupting data when we have chipset x and graphics controller (!) y and Windows XP SP II OEM, but now with any other combination? This user says that fonts are too small, this other user says that the fonts are unnecessarily large - well, maybe we should make them adjustable, except that oops, our window layout broke.
To be honest, I've never needed any graph theory. The spirit of the Big O notation is good to know and acknowledging floating point issues is mandatory, but other than that, it's better to just try out (prototype) the algorithm and measure it rather than try some meticulous analysis on paper. Virtually all solvable algebraic problems have been solved in off-the-shelf math library packages, so why should I know how to implement them? All I need to know is that it's a black box, I put something in and get something out, and don't forget that there are limitations, so always check that the output makes sense, and if needed, find the edge cases where it fails. Again, by prototyping. (If you want to sound more scientific, call it "simulation".)
What is essential is the skill to find what has already been made and understand how to make use of it, instead of wasting your time on re-inventing solutions to problems that already have been adequately solved. Stand on the shoulders of giants, think hard and try things out.
Addition: I'm of course biased towards what I do (as we all are), so don't take my word for it. According to user SK-logic in comment below, the following positions require knowing advanced theoretical concepts: game dev., engineering graphics dev. (CAD/CAE/...), compilers, finacial math (high freq. trading and such), database engines, complex business logic (e.g., in logistics), AI and NLP (machine learning, etc.), and many many more.