The pronunciation of polyphonic character is usually deduced from the semantic of the context. In a transliterated word, there is no such context, the character only stands for its pronunciation, so it seems not a good idea to use polyphonic characters for that purpose.
However they are used in transliteration. For example 什 is extensively used for Russian names. More interestingly, there seems to be an agreement that in transliteration, its less common pronunciation shi2 is always used instead of the more common shen2.
So my questions are, in transliteration field,
- Is there any guidelines around/against using polyphonic characters?
- How does one decide which pronunciation is the 'agreed' one to use?
- Is there any polyphonic character, of which multiple pronunciations are used for transliteration?

shen2, the actual character is甚as in甚麼. The Middle Chinese pronunciation of 什 isdzyip, and there was no -n or -m ending pronunciation. Still in Korean and Vietnamese today, 什 has only one pronounced with -p ending. The natural descendant of this pronunciation in today's mandarin is `shi2'. – user58955 Oct 08 '13 at 22:23jiaois 白读 andjue文读; for 薄,baois 白读 andbo文读. Which is which concerns Middle Chinese pronunciation. Usually 文读音 is used in translation, that's why Berlin is 柏林 (bo2 lin2) although the usually pronunciation of 柏 (the name of the tree) isbai3. – user58955 Oct 08 '13 at 22:29