Ed Popko, you wrote:
"Venus was not so sharp that I tell you the exact phase. Perhaps I was measuring her center and not even aware of it."
Yes, that's what you're supposed to do. That's what the historical sources recommend. And that's what you learned in my lunars class. Indeed, if you were using a "limb" of Venus, then all of your sights are off by a few tenths of a minute of arc, right?
And you wrote:
"The technique I'm using lately is to keep one body dead center in the scope and sweep the other one back and forth across it. The 'trace' of the sweeping body in my mind's eye is what I draw on to bring them in kissing distance."
Yes, that's exactly what you're supposed to do. This is also how you get best contact with artificial horizon sights. This technique for lunars only fails when the angle is very short (which are of no use for time/longitude but interesting in their own right). With short distances, "sweeping" has hardly any effect. This isn't a problem --just a difference.
"When I feel I have it, I stop sweeping, hold it to check the final position and quickly record the time (pause Frank's GPS Anti Spoofing app)."
It's certainly entertaining that you're using my app as a GMT/UT stopwatch. Nothing wrong that, of course! One note though: you say you record the time "quickly". Don't forget that one nice feature of lunars is that you can relax about recording the exact time. The distance changes by a tenth of a minute in twelve seconds of time. There's nothing like the rapidly-changing angles that we deal with in ordinary altitude sights.
Yet another way you could use my app. It's supplying the body altitudes! Before you start shooting your lunars, "record" a Moon altitude sight and then a Venus altitude sight from the app (standard "screen cap" function on your device will save you some writer's cramp). Then shoot four lunars. Then do another Venus and Moon simulated altitude in reverse order. Now average them all. That is, average the four lunars, and separately average the Venus altitude pair and the Moon altitude pair. The goal here is to get a set of traditional observations for a complete lunar analysis when you're not at a location with a horizon.
And you wrote:
"For myself, I can only do lunars when supported by something - my elbows resting on the hood of my car, shoulders propped against the corner of a building, sitting in a chair etc. If I simply stand and free-hand the shot, the results are always poor."
Hold the sextant frame with both hands, one on each side. Do your sweeping, look at the gap or overlap, and then lower the sextant from your eye. Give the micrometer a little tweak. Then look again with both hands on the frame. Again, as noted above, lunars are relaxing sights. You do not need your fingers on the micrometer, chasing the correct angle, as is normal with altitude sights.
"With the Astra, however, I have to study the vernier very closely. Unlkie the higher quality C&P vernier which is markd every tenth of a minute, the Astra is only every two tenths. So I visually interpolate the result ot a single tenth of a minute if two vernier marks fall between two frame marks (a Stan Klein trick)."
You do not need the micrometer vernier at all. It's nothing but marketing... "Oooo, gosh golly, this sextant micrometer is graduated to tenths so it's super-duper accurate!" No... You can read the micrometer to tenths just by looking at the primary pointer. Eyeball it. That little vernier adds nothing. Plath sextants from the 1950s --often considered among the best ever made-- did not have verniers on the micrometer. They are un-necessary.
You mentioned in your reply to Andres that you were intrigued by the way that your error seemed to decrease over time. This is almost certainly just the accident of statistical noise on this one round of sights. It's true that the very first lunar you take in any run is often the worst, and I consider it reasonable practice to just ignore the first one. Don't even record it. After that, you will sometimes see patterns in your results, but if you're looking at only a handful of sights, the patterns are likely to be in your head as much as in the data. The catch is that when we start paying too much attention to these illusory trends, we end up imagining accuracy as well as inaccuracy that are not real properties of the data.
Finally, always average sets of four lunars. The individual sights are data input. It's of no real consequence how you perform on each and every observation. Think of a completed lunar as being the average of a set of four. Averaging sets of four not only reduces random error by a factor of 2 (square root of the number of sights in general), it also cancels your concern about reading the vernier. Suppose you have a set of four sights diligently recorded to the nearest tenth of a minute. You average those. Now take the same sights and record them to the nearest two tenths of a minute and average those numbers. You will find that your two averages are identical to the nearest tenth three out of four times on average. In 25% of cases they differ by +/-0.1, an inconsequential difference.
Be sure to keep records of lunars binned by the average lunar angle. The best use of lunars in the modern world is to generate your own arc error table. If you consistently find average lunar errors near zero at, let's say, 30° and 60° distance, but you also consistently find errors of +0.3' around 45°, then you have found good evidence that your sextant has a modest arc error in the vicinity of 45°. Consistent errors are correctable errors. Collect enough data, and you then have known errors that you can apply to all future observations.