Tony,

First forget about compression mechanisms. The same compression mechanism will work with all protocols as long as the protocol can send true binary data (no special codes)  The amount of compression is highly dependent on the entropy of the source data.  In Winlink for example ALL protocols are sent using the standard FBB B1 compression mechanism (used since the 1980's) packaged in a B2 envelope that allows multiple files per message.  This is independent of protocol used.

The theoretical maximum data rate of a protocol itself doesn't mean much.....E.g. it is straightforward to develop a high throughput mode (e.g. 256QAM that sends one byte/symbol)l but unless the signal is strong and the propagation has minimal fading and minimal multpath it will not be usable in practice.  So the best way to compare modes or protocols is this:

1) Use a HF channel simulator that can model specific S:N across a 3 KHz bandswidth.  The simulator is statistically repeatable so allows measurements to be duplicated and confirmed INDEPENDENT of changing HF conditions. 

2) The simulator must be able to model the important standard channels used in HF propagation modeling. E.g.  Multipath quiet, multipath moderate, multipath severe.  These are well documented and used in industry.

3) Run a real ARQ connection using large blocks of typical data using ARQ sessions of at least 5 minutes duration. over the standard HF channels above at S:N (3KHz) levels of -5 to +20 dB.  Compute the NET (after all repeats, mode shifts etc) ARQ throughput.  Repeat the measurements using longer times if necessary until repeatable results (throughput vs channel type @ given S:N) is obtained.

On HF transmission it is rare to see channels much better than 10-15 dB S:N and multipath quiet.  More often we have signals of 0 to 5 dB S:N and multipath moderate or worse.  Real world performance prediction is to use these more typical conditions to evaluate and compare protocols.

I have a MSEE degree in Computer and Communications have been working with HF protocols now for over 20 years and have come to understand that casual over the air comparisons or comparing protocols based on just their specs alone is not usually productive and not repeatable.

VARA does look to be a promising protocol and newer high symbol rate modes like P4 and STANAG /Mil STD 188 variations can achieve high throughput in challenging conditions.  What we should be doing is pushing for using updated rules (like regulations by protocol bandwidth and not symbol rate/carrier)    which will encourage more DSP development for faster and more robust HF and VHF data transmissions.

73, Rick Muething, KN6KB