Damn, I just wrote a long reply to this, but Listserv decided that I'm not 
allowed to write for that long and asked me to re-login, deleting the draft in 
the process. Grumble... Okay, synopsis:

1) Since you have prenasalized versions of pretty much everything, I think 
analyzing these as clusters all the way would make better sense than 
analyzing some as clusters and some as phonemes.
1.1) You seemed to lack prenasalized aspirates, so consider: n+th = [nt], n+t 
= [nd].
1.2) Gradation-wise, how would ranta<>randan, randa<>rannan sound like? 
Would make more sense than ranta<>rannan, randa<>randan IMO.

2) Completely lacking palatalized/labialized aspirates in a system this big is odd.

3) Contrasting /c c_j/ is extremely bizarre. /c cj/ could do tho, or 
palatalization affecting the vowels substantially (a la /ca/ = [ca], /c_ja/ = 
[cE]) but you seem to have too many of 'em for that to work.

4) Contrasting /s S s_j S_j/ is also odd. /s S_j/ and /s s` s_j/ ~ /s S S_j/ are 
more typical systems.

5) Having [nD_j] but no [D_j] is a bit odd too, but less odd than the others.

6) I look forwards to more of this project!

John Vertical