Directional dependency of connection off RF signal generator BNC to a resistive load

A BNC cable is simply a coaxial cable with a solid center core and a outer conductor sheath right? The signal coming out the output BNC of standard signal generators is the center conducting core carries the potential while the outside conductor sheath is always grounded?

Can someone explain what the theory is behind the following then:

Imagine a setup with two resistors connected in series. One resistor is low resistance (50ohm) and another is high resistance (1Mohm). The generator has standard 50ohm output impedance so clearly impedance mismatch with the generator. We use a 1MHz signal at say 10Vp-p and zero DC. We take the signal out of the BNC cable and wire it such that the 'hot' end and the 'ground' can be interchanged with respect to the terminated ends of the series resistor. There are two possible configurations: (1) Hot end at high resistor end and ground at low resistor end and (2) hot end at low resistor end and ground at high resistor end.

Imagine I have a way of determining if my resistors are being delivered any power (small resistor is of interest). Say I was measuring the temperature of the smaller resistor to determine joule heating (thus related to power drop to small resistor). Why is it that the resistor gets delivered power under configuration (2) described above, but does not do so for configuration (1) (to any appreciable degree at least)? In other words, having the high resistance load in front of the signal coming off the signal generator 'hot' terminal, effectively blocks power delivery to the smaller resistor?

So even for half the period of 1MHz excitation where the potential is negative typically in DC may indicate opposing current flow, it should then as logic may dictate to eliminate this 'directional dependency', but this doesnt happen. There is still a directional dependency of where the actual 'hot' end is placed. Is there a reason for this?

As long as the resistors can be described as lumped elements (circuit dimensions << wavelength, which is 300 m for 1 MHz), the order doesn't matter.

A real circuit will also show parasitic capacitances that should be added to an equivant circuit to get a realistic picture. If you trying to measure voltages with a real instrument, e.g. an oscilloscope probe, the probe impedance has to be considered, too.