ISCII is great to store names(people/location etc) which do not vary across languages. Consider a 10 Million names database storing names of people which need to be picked up during reports across different languages. One row storage of ISCII can take care of the names and transliteration provided by .Net encoding classes (similar effort can be applied to Unicode too but without lot of success) help display in various indic languages. In case Unicode encoding you will need to store a language specific name( this too could be useful if you are hell bent on correcting names/matras to suit local language) -thus multiplying storage cost.
The cost of storing ISCII is offset by need for translation into Unicode for display(IE is quite ahead in terms of display of unicode data with appropriate font)/capture (with help of INSCRIPT or local variant of phonetic or web based entry).
Indexing/sorting – language specific sorting can be different (Tamil is very different from other languages). A topic for another post alltogether.