derivation help?

heya!

    So, I decided to go ahead with things, and am developing a "new" celt=
ic lang
without the meld with English. What I've been doing is taking Old irish w=
ords,
deriving them in my head to what I think they should be, and then backtra=
cking to
how I got there. Very much an analeptic way of doing things (or is that p=
roleptic?
read Bohm to straighten it out). Anyway, I've come across a couble glitch=
es I
thought you folks might be able to help me with. I'll start with some sta=
ndard
transformation and the rules I found, and then go to the ones I didn't kn=
ow how they
got there. Any ideas on possible derivational rules would be greatly appr=
eciated!

    OI.=3DOld Irish    A.=3DAelya

    OI. odb 'knot in wood' > A. ydh
       -svarabhakti (OI. db pronounced /Dv/ >/Div/)
       -loss of final C (in polysyll)
       -i-mutation (o>y)
       -loss of final V (in polysyll)

    OI.salann 'salt' > A. sal
        - simplification of final C or cluster
        - loss of final C
        - loss of final V

    OI. fedb 'widow' >  A. feidh (dh=3D/D/)
        -svarabhakti as above
        -loss of final C (/feDiv/>/feDi/)
        -i mutation (e>ei)
        -loss of final V

    OI. sen 'old' >  A. syen/sien (can't decide which orthography I like =
better)
        - just that s before front vowel =3D /S/

    OI. dobur 'water' > A. d=FAr (u with acute accent =3D long)
        - b>w/V_V
        - simplification of vowel cluster (owu>=FA)
        - loss of finals would then happen, but word in now monosyllabic,=
 so no
change)

    Okay, that should be enough to give the idea. On to my questions:

    OI.  an=E1l (anaal) 'breath' > A. an
        -loss of finals as normal
                 But what about that long a? Only short vowels are lost n=
ormally.
Maybe an earlier rule that shortens long vowels in the final, nonacc syll=
able? Of
course, all non-initial syll are nonaccented, so...

    OI. bendacht 'blessing' > A. bennaeth or benn=E1th (long a)
        - ch /X/ aspirates following plosive
        - voiceless aspirates fricativized (/th/>/T/)
        - vocalization of /X/ (as in Old English)
        - diphthongization of V (a>ae) OR lengthening of V (a>aa/=E1)
        which makes more sense in terms of treatment of that vowel? and h=
ow come the
finals weren't lost? Which derivation do you like more?

    OI. scethach 'emetic' > A. syetha/sietha (same problem as above)
         -sc>sy/#_
        -loss of final C
    why no loss of final V?


    And one additional question : wouldn't this loss of finals cause an a=
wful lot of
monosyllabic and homonymous words?

    Any ideas are appreciated!!

    Clint