Well if you have a copy of Johnson's "High Speed Digital Design: A Handbook of Black Magic", and if you don't you should :) look at page 192. Here he shows a formula you can use for the ratio of crosstalk for a given input voltage.
Below that is an image showing that D, the distance between the aggressor and the victim, is measured center to center.
The Ultracad calculator references two Ultracad articles that in turn reference Johnson's book (and the above formula) directly so that is where they got their equations for their calculator.
Looking at the bottom of the EEWeb calculator page you will see they have a formula like this:
Vcrosstalk = V * 1/(1+(S/H)^2)
You might recognize that as the same ratio equation used by Johnson and Ultracad.
My conclusion would be that they have drawn their picture incorrectly but you are getting the same results because the underlying formulas are the same. This would not be the first time I tried an EEWEB calculator and got inaccurate results or bad information...
If you want to experiment more with crosstalk and can get your hands on a simulator like Hyperlynx, Sigrity, your favorite simulator, that can be a intuitive way of understanding things.
Finally my favorite all time experiment for crosstalk is just to get a piece of copper clad board, two wires, a signal source, and an oscilloscope. Lay the wires on top of the board and terminate the far end of both wires with a resistor. Connect the other end of the second wire to the top of the copper board. Then using the copper board as your gnd (reference plane), run a signal through the remaining wire. You can then measure the crosstalk voltage on the first wire with your scope. You can see what happens as you move them further apart or closer together.