Avevo in mente anche un valore compreso tra 16° e 17° Est, e in pratica nessuno dei due è davvero sbagliato.

Se ricordo bene, il valore più preciso è intorno ai 16,3°, quindi si tratta semplicemente di un arrotondamento. Seguendo il tuo ragionamento (che il cifrario dovrebbe essere decifrato “manualmente”) come avrebbe potuto fare lo stesso Zodiac, quella differenza di un grado è a mio parere facilmente trascurabile, dato il livello di precisione della mappa.

È inoltre opportuno tenere presente che la croce sul Monte Diablo è piuttosto irregolare, quindi quel grado in più può essere ragionevolmente regolato in entrambe le direzioni senza influenzare significativamente il risultato finale.

Posted by: @shaqmeister

↑

This takes us, for this example, to somewhere in the region of West Portal in San Francisco, roughly.

For whatever reason, likely simple carelessness, the scaling was not accurately deduced when rendering this first example. With the appropriate degree of precision applied, we should have arrived at the following for “at four and a third inches, radian eight”:

As an alternative to plotting candidate solves manually using some such unsuited application as Microsoft Word, if your skill-set permits it, it can be helpful to hack together a script (Python, for example) to do the plotting for you. I have been recently playing around with just this approach using matplotlib in a Jupyter Notebook. This has led to some success, although the coding, I can be sure, is not pretty, nor likely efficient. I have, all the same, been able to set up and calibrate a polar coordinate system over the Phillips 66 and position this so that it centres on Diablo, is re-oriented with 0 at the top and angles going clockwise, and scaled appropriately to inches on the map whilst accounting also for the magnetic declination. With all this in place, it is then straightforward and quick just to plot candidates via their coordinates, and display them on the map along with labels. 

Here’s the output from my most recent example, which took a number of random candidate solutions and applied a straight-to-map plotting to them.