It's because a voltage source has an impedance or resistance of zero. So for impedance analysis you can set it to zero.
Thevenin's theorem also states that you can replace a voltage source in series with a resistor with a current source in parallel with that resistance (or vice versa) - and everything works the same.
There is a Wiki article (and lots of other stuff on the web Google Thevenin Theorem and Norton Theorem and Kirchoff for the full set of key linear circuit theorems).
But you are correct - the resitors are NOT in parallel, it is that the combination of voltage source and resistors can replaced by a single resistor (value = the two in parallel) and a voltage source (which will have a different voltage from the original source).
I can't draw diagrams from here but there are lots on the web.