Tom,
I did study the problem of dish gain but have not made any actual measurements of the TMFS instances. For this work I have used three references: (1) Antennas, Kraus and Marhefka, McGraw-Hill, 2002, 3rd edition; (2) Principles of Communications Satellites, Gordon and Morgan, John Wiley and Sons, 1993 and (3) The ARRL Antenna Book, ARRL, 2000, 19th edition. These three references provide a comprehensive picture of the issues surrounding this problem.
In short system gain is given by the dish diameter and surface roughness relative to the operating frequency and modified by the feed's illumination pattern (so-called "gain taper" towards the dish edge) which yields the "efficiency". The attached figure (from Kraus and Marhefka) shows the trends for a 50% efficient system which is representative. There is a curve for a 20 m diameter dish which is close enough to our 18 m. At 1 GHz I read roughly 43 dBi and in agreement with the previously mentioned value (BTW I wouldn't use anything more significant than the units place for this work). Scaling to 433 MHz (inverse wavelength squared) yields roughly 36 dBi. Again, in agreement with your previous data.
The surface roughness is less significant for these lower frequencies so I suspect that little correction is required for this effect. Further, a 50% efficiency is very reasonable from a design perspective (this corresponds to a roughly 10 dB edge taper suitable for effective "spill-over" management) and actual achievable performance (don't forget to take the feed gain into account when putting together the link budget).
----------------------------------------
One additional thought: it is possible that the surface roughness of the dish is large enough to begin impacting the gain at frequencies at 2 GHz and above. Referring to the attached figure it can be seen that a surface roughness of ~0.5" (rms) will cause the gain to peak at a frequency of ~1.8 GHz. I have looked over the upper dish surface to get a sense of its quality and believe (no data to back it up) that it has got to be at least as good as this in terms of roughness although I cannot asses what the opticians call the overall "figure" or the adherence to a parabolic shape. Something to consider at higher frequencies.
Best regards to all,
Joseph
For more information see dish gain calculations by John Ewan