Translation's doubts

Although a good part of my time is spent into reading and translating Sanskrit texts, I still never faced the issue independently of a given sample. The problems I most frequently encounter are:

1. a Sanskrit text lacks punctuation, footnotes, italics, brackets, etc. Hence, a Sankrit-speaking (or Sanskrit-writing) author has to express in words what we would express through the above devises. But the border between 'words' and 'punctuation' or between 'text' and 'footnotes' is not always easy to understand.

2. often, a single English (or Italian) term seems to render various Sanskrit ones, e.g. 'to produce' (kṛ-, sampad-, prasū-, jān-, utpād-, kalp-…). 

3. often, a single Sanskrit term can to be translated by many English (or Italian) ones, e.g., eva ('very', 'just', 'precisely', 'verily', 'alone', 'only'…, notwithstanding the case in which it only indicates that emphasis is to be added to the preceding word, see n.1 above). 

4. sometimes, two Sanskrit terms can be translated with two different English (or Italian) ones, but the author may not be consistent in its use, so that they might at times coalesce into a single meaning (apekṣ- and ākāṅkṣ-, for instance, may at times by synonyms, at time not). An extreme case is karman, which may have several meanings and is at times just a synonym of kriyā. 

I do not mean to say that the same word has to be translated always in the same way. On the other hand, a single Sanskrit (English, Italian, German…) word can have different meanings, which have to be rendered in the target language by different terms. I am only claiming that as long as the meaning is the same, it would be preferable to translate it with just one and the same term in the target language, in order to convey at least a bit of the semantic richness of the term and also in order to make life easier for readers who know (a little bit of) Sanskrit and can easily think at the Sanskrit equivalent. 

4. puzzles me. Should I translate such terms in the same way or should I use different English terms, leaving to the reader the burden of understanding that they are the same, but also the chance to appreciate that, still, there is a shade of meaning I might have overlooked?

The problem boils down to this: should the translator leave the text as it is (with all its inconsistencies and ambiguities) or not? The question is easily answered by translators of Western philosophers (I happened to discuss it with a colleague who translated E. Bloch, Micaela Latini), but in their case reader and writer share much more.