Strictly speaking, the output impedance of most amateur transmitters is not specified. The specifications instead state what the output power of the transmitter will be into a 50 ohm load. Often the reader will infer that this is the condition of impedance matching or a conjugate match. While a conjugate match is the condition of maximum power transfer, it is not necessarily the condition of optimum efficiency for the transmitter. It is quite common that the manufacturer of the transmitter has specified a load impedance that improves the efficiency of the transmitter in order to reduce the power supply requirements or the heat generated by the transmitter final amplifier.
With regard to the SWR question, the output impedance of the source (the transmitter in this case) does not enter into the equation for SWR which is simply based on the relationship of the load impedance (the antenna in this case) to the characteristic impedance of the transmission line.
When a source (transmitter) initially puts power into the transmission line, it is not "aware" of any mismatch of the load to the transmission line. If the load matches the characteristic impedance of the transmission line, no power is reflected so the transmitter simply sees the characteristic impedance of the transmission line. If the load impedance is not a match for the transmission line then some part of the power is reflected. The amount of power reflected is equal to the reflection coefficient squared. So for a 2:1 SWR for example, the reflection coefficient is 0.333. Square that and you find that 11.1% of the power arriving at the load will be reflected back toward the source (ignoring the losses in the transmission line for now). The power returning toward the source has the effect of altering the impedance that the source "sees" due to the forward and reflected waves combining/interacting. The exact impedance that the source sees is a function of the SWR, the frequency, and the transmission line.
For example, 50 feet of Times LMR-400 on 14 MHz with a 2:1 load will present an impedance to the source (transmitter) of 33.5 +j21 or a Z of 39.5 ohms due to the reflection. The source doesn't see this new impedance until the first reflected wave has returned to the transmitter. Of course this happens at 66% to 90% at the speed of light so the resulting change in impedance occurs seemingly instantaneously.
Tubed transmitters, due to their tunable output circuits through front panel adjustments, have the ability to have the output impedance of the tube final transformed to the impedance seen at the output connector. Thus once the reflected wave has altered the impedance that the transmitter sees, the operator can adjust the matching circuit to compensate, within a limited range, for the altered load impedance presented by the transmission line. This capability to alter the impedance matching circuit is functionally equivalent to the role of an external antenna tuner used with solid state transmitters.
A 12 volt solid state transmitter with an output power of more than a few watts, will have a final transistor amplifier circuit with a very low output impedance. In order to drive a much higher impedance 50 ohm load, the final output must be transformed through the use of an RF transformer or other suitable impedance transformation method. Since there are no operator adjustments in this output circuit like there is with a tubed final, any changes to the load impedance will alter the current through or voltage across the final transistor. Thus solid state transmitters have power foldback circuits that kick in around a 2:1 SWR in order to prevent the transistor from overheating, reaching breakdown voltage, or exceeding the maximum allowable current for the device.
Before coming back to the question of the transmitter returning the reflected power, we first need to account for the very real effect of transmission line loss. For every trip that the power takes along the length of the transmission line, some percent is lost. It is this back and forth movement of power along the transmission line that accounts for the extra loss due to a mismatched load. In the example above, each trip reduces the power by another 5.25% (-0.234 dB). So with a 100 watt transmitter, the power that reaches the load on the first trip is 94.75 watts (100 watts * 0.9475). 11.1% of this power (10.5 watts) is reflected back toward the transmitter. But this trip also experiences loss so the power returned at the transmitter is 9.97 watts (94.75 * 0.111 * 0.9475). All of these losses result in heat (the majority of the losses on HF bands are due to copper losses). We can now see that the transmitter NEVER returns all of the power reflected by the load since some of it is burned up as heat.
Does a transmitter return all available reflected power to the load? The answer is yes provided the changed output impedance presented to the transmitter (due to the reflected power) does not alter the output power of the transmitter. The power that the transmitter returns toward the load will, in the above example, again have about 10% (11.1% * 0.9575 * 0.9575) returned as measured at the transmitter. This ping-pong effect continues until all of the power is transferred or burned up as heat. Generally, within 4 to 10 trips, the amount of power bouncing back and forth has become vanishingly small.
We can empirically prove that the re-reflections by the transmitter are taking place by calculating the infinite geometric series of the losses occurring from the theoretical bouncing back and forth of the waves due to the multiple reflections and then comparing this to the actual output power of the antenna and the output power developed by the transmitter.
It is a common misconception that the output impedance of a solid state transmitter is the same as the impedance "looking into" the transmitter. As an active RF signal generator, this is rarely the case. In the lab, this condition can be met by using an attenuator of 20 dB on the output or more of the RF source. This effectively isolates the RF source from the reflected power and establishes the condition that the RF generator outputs a nearly constant RF voltage regardless of its load conditions.
The output power of a solid state transmitter will be reduced by an amount exactly equal to the power reflected from the load due to the change of input impedance of the transmission line. The reflected power is added to the output of the now reduced output power of the transmitter and sent toward the antenna (load). This meets the required condition that incident power is equal to the source power plus the reflected power. It is notable that the power output of the transmitter, before the reflected power has returned to the transmitter, is the same as the total power sent toward the antenna once the reflected power has reached the transmitter (thereby altering the input impedance of the transmission line and reducing the transmitter's output power).