Jeff, your re-reflection concept is complicating things.

Just consider a typical ham FM voice transmitter after steady state has
been substantially established.

If the load was for example 70+j0 and the transmitter was connected by 2
wavelengths of Belden 9258, accounting for line loss we would expect the
transmitter to see a load of about 68+j0 (VSWR(50)=1.4), ie the ratio of
v/i at the output terminals of the transmitter would be about 68+j0.

Consider also the case where the transmitter was connected by 2.25
wavelengths of Belden 9258, accounting for line loss we would expect the
transmitter to see a load of about 37+j0 (VSWR(50)=1.4), ie the ratio of
v/i at the output terminals of the transmitter would be about 37+j0.

If you were to measure the output power of a range of such transmitters,
it is unlikely that they will produce substantially identical power
output under both conditions though the VSWR is similar, that the power
is changed from that with a 50+j0 load by the same amount in all cases,
or that the change is reliably predicted by your analysis technique
(based on VSWR).

Your treatment of the 'forward power' as a constant with different loads,
and the approximation of transmisssion line behaviour contribute error.

Additionally, the formulae you use do not account for non linear
behaviour of typical output stages, gain variation at different output
level, whether they reach voltage or current saturation with a given
load, the effects of PA protection schemes that might limit current,
'relected power', 'power output' etc.

I would agree that a 70 ohm antenna at the end of 4m or so of RG8/X will
*probably* not result in a large loss of output power, but I wouldn't
agree with your results or method because it is not sound. Nevertheless,
I understand why Ed might want to transform the load to 50 ohms, and he
shouldn't be discouraged by flawed estimates.

Owen

PS:
You might dismiss my example of the 5 and 500 ohm loads on a short line
as unrealistic, but it exposes a common misunderstanding that the loss
per unit length when VSWR1 is uniform along the line. If you want to
explore the idea further, I have written some notes at
http://www.vk1od.net/VSWR/displacement.htm .