I'm not sure about those, but I am a huge fan of the late, great Peter Connolly, who was an illustrator of classical civilization and mythology. I don't always agree with his reconstruction but he always shows his work and justification.

For your second question, you are right in thinking there are some problems. Sometimes it can be somewhat straightforward, when the roof tiles are terra cotta, for example. It can still be laborious to measure and analyze the tile remains to make an accurate reconstruction (a bit like a jigsaw puzzle), but it can be done. When the tiles aren't made of something durable like terra cotta, however, it is usually impossible to tell whether it is thatch, wooden, open sky, or what have you. This can have significant effects on broader interpretations as well: how you imagine Celtic roofs, for example, significantly impacts how you imagine the level of elite material prosperity.

And the backprojection you mention is another issue, and one that can sometimes be taken rather too far. For example, you will often see reconstructions of the Shang palace at Yin look like something straight out of the Forbidden City, even though there really isn't much in the way of reason to think Chinese architecture remained that static. This also carries an ideological bent--by making Yin palace look that "Chinese", it cements the traditional narrative of the Shang as the "origin" of Chinese culture and thus creates an unbroken "China" stretching back to the Bronze Age.

But all reconstructions have these problems to greater or lesser extent. The good illustrators are uprfront about this and clearly explain their reasoning.

A map, rather than a painting, of Tenochtitlan was made in 1524 and published along with a Cortes' second letter by Friedrich Peypus in Praeclara de Nova maris Oceani Hyspania Narratio.^1 The Nuremberg Map does not, of course, depict the city at the time -- it was still being rebuilt from the devastation of the months long siege in 1521. What it does show is a 16th Century European cartographic take on the city at its height.

Sometimes attributed directly to Cortes -- who cannot be ruled out as the creator -- the actual artist is unknown. The earliness of the maps and the faithful detail it provides of the Templo Mayor district, does indicate that it that it was made by someone who either had intimate knowledge of Tenochtitlan, or was working from an earlier source made by someone who did. Mundy (1998) even notes the map has a common Mesoamerican orientation -- West at the top -- leading to some speculation about a possible indigenous author.

The map itself adopts very European styles in depicting the rest of the city, so there isn't much to infer from the its depiction of the rest of the city, other than its density. We can though, see the network of canals that ran through the city as well as the placement of the causeways and the neighboring cities, all of which are attested to in other sources and are ubiquitous in modern maps and models of Tenochtitlan.

^(1 The full title, translated from the Latin in which the book was published is, The splendid narrative of Ferdinand Cortes about the New Spain of the Ocean Sea, transmitted to the most sacred and invincible, always august Charles, Emperor of the Romans, King of the Spaniards, in the year of the Lord 1520; in which is contained many things worthy of knowledge and admiration about the excellent cities of their provinces. . . above all about the famous city of Temixtitan and its diverse wonders, which will wondrously please the reader.)

The series Time Team made extensive use of an artist, Victor Ambrus, to help show viewers what the site they were digging would have looked like.

He has a book - http://www.amazon.com/Recreating-Past-Victor-Ambrus/dp/0752419099 - which may be of interest. British archeology, though - I very much doubt there will be any Incas or Aztecs!