Poor fit is relegated to non-significant areas. Radial distortion is minimised by only using the central portion of photographs whenever possible. Precision of surveyed data points and the fit of pics to them is thus generally better than 1 minute of arc, the limit of precision for normal eyesight. Poorer precision may occasionally occur where surveyed points are >5 ° apart and / or particular surveyed points are just naturally incapable of fine resolution. Obscured or unsurveyed sections of horizon may have an approximation of the hidden horizon drawn as a green line. These green lines are usually based on / informed by horizon data generated by (historically) HeyWhatsThat.com and / or (more recently) the "Horizon" computer program (using NASA's SRTM 30m dataset [downloadable here]). These green lines are indicative but should not be regarded as accurate because the base dataset is insufficiently precise.
Scale is one unit = one degree. Minor ticks are one degree long, major ticks two. Labels for the solstitial eight day brackets and for the sixteenths adjacent to the lunar limits have been omitted to reduce overcrowding but their trajectories are there. Dashed trajectories are lunar centres for the thirtysecondth divisions, in other words the boundaries between the parts centred on the sixteenths. Luni-solar discs are 0.52° in diameter and are (with naked horizons) centred 0.5° above the horizon, thus they may be used to estimate the effect of possible prehistoric tree cover.
Good pictures of distant horizons are difficult to get, due to atmospheric haze. Naked eye is best for overall clarity, the camera is worst and the theodolite is in the middle. The oldest photos are scans of 35mm slides but from 2007 all have been digital. Since then the number of pictures taken has steadily increased but it has taken a while to develop a full panoramic technique. Originally, key horizon segments were photographed, often with the monument in the foreground to give context. Now, as well as views of the monument, standard practice is to take a full 360° sequence of overlapping shots with a short lens and then another set with a longer lens. Only at about 300mm does the camera reach the same precision as the theodolite but this is rarely necessary. Standard lenses have been used, so there has been no attempt to produce seamless panoramas. These pictures are cylindrical projections, precisely accurate at the horizon.
Do bear in mind, when looking at the pictures, that they give a fixed view but in the real world one can shift the position of the lunar / solar disc against the horizon (if it is not too distant) by changing observer position. Survey position generally attempts to be at or near the centre of the monument. For various reasons, the photographs may not all be taken from the same place. The final image is likely to be a patchwork of photos taken with different focal lengths, from different viewpoints and at different times, but even so the horizon has always been accurately fitted to the survey data.
Pictures may contain views of the monument itself to give an impression of its location and orientation. However, the horizon is still configured as seen from the survey position and not from the viewpoint that includes the monument itself. The monument image has been inserted by matching two key horizon points with those of the surveyed horizon.
In the early days, individual photos were annotated using graphical software. Abbreviations indicating the measured declinations, or named luni-solar calendrical events were pointed at the surveyed horizon points. Interpolation was not possible. The ability to accurately scale images has in turn lead to changes in survey technique. When a natural marker lacked precise definition, I used to estimate the visually optimal position as a survey point. Now I would, in preference, chose specific and accurately notable points of sufficient size and definition (houses / bushes etc) in the general vicinity, so as to enable the photos to be fitted to the data more precisely. This means that the pictures now give a better model of how the horizons were being used. The price for that has been fewer "accurate hits" in the survey data tables.
Be aware that the date chosen to generate the trajectories may not be the same as the one used in the data tables. The standard reference date for the data tables is 1800BCE though I have used different ones for some of the older sites. Although slightly early for the Cork-Kerry monuments, 1800BCE was originally chosen because that was the year Thom used in his work. Changes in the obliquity of the ecliptic at that time were round about 0.01° per 100 years in either direction which is not hugely significant when one is only looking for general correlations. However, the more I've done, the more I've realised that they were very good at this, very accurate, and that some sites were chosen to give really good precision to particular events. Often related lunar and solar events. So when I do the trajectories I generally choose a date that is likely to be closer to the actual construction date. For the Cork-Kerry Bronze Age sites this is usually 1200BCE.