It's tough to tell precisely what you're after. Perhaps you could give an example of what you're looking for.
My first response is to recommend that you look into R http://www.r-project.org/ and some of the GUIs based on it. It's quickly becoming a standard in any field requiring statistical analysis, from quantitative finance, to medicine, to social sciences.
It has a hideous programming language but a huge package archive http://cran.r-project.org/. I would recommend poking around the statistics stack exchange site for pointers, but some notable GUIs include RapidMiner (http://rapid-i.com/content/view/181/190/), RStudio (New and just getting out of the gate, so don't expect a lot right now http://www.rstudio.org/), Red-R (http://www.red-r.org/), R Commander (http://socserv.mcmaster.ca/jfox/Misc/Rcmdr/), and Rattle (http://rattle.togaware.com/). R is also integrated in SAGE, with which you may already be familiar. One huge benefit of R is Sweave, which allows integration with LaTeX, so your papers can render your data and publication at the same time.
If you just want a graphing solution, WaveMetrics IGOR (http://www.wavemetrics.com/) is probably the best bet. It's available on Windows and Mac OS so you can share files with other people fairly easily.
KaleidaGraph (http://www.synergy.com/) and pro Fit (http://www.quansoft.com/) are notable options as well.
One of the secrets you'll find out when hanging out on various visualization forums, especially Edward Tufte's (http://www.edwardtufte.com/bboard/q-and-a?topic_id=1), is that a large number of the published graphs (especially his) often start in Excel and are exported to be manipulated in Adobe Illustrator. You may have a graph in mind that comes from this pipeline, which has a large manual effort but a magnificent potential output. I've also done similar with Mathematica (my tool of choice) when needing print-quality output.