For scientific plotting with lots of features, PGPLOT is still the best that I have seen. I use R, also. PDL has some interesting graphs. Gnuplot covers most of the simple cases, and solves a few hard problems, also. Take a look at Nutmeg if you have complex numbers.

Automatic generation of Really Nice Graphs in the General Case is still an Open Problem. It doesn't really matter how much money you spend. You want to paint a portrait of your data. It's art. There is taste involved. The charts at the Dundas link are extraordinarily ugly to my eye, for example. To me they look like 'how to lie with statistics' examples for the marketing department. I prefer the graphs with a more scientific look, such as these PGPlot examples. If you want good looking output, use the Postscript driver and convert to whatever you need.

On the other hand, there are some very nice algorithms for autoscaling, etc, that are missing or poorly implemented in the 'widely used' graphing program.