Joe Hill wrote:

If PIE /k_w/ turns to PG /x_w/ or /h_w/, surely, PIE /k_wetwores/ should come out as /x_wiTwores/ instead of /fiTwores/ Or am I completely wrong?
No, you're correct-- the initial *kw _should_ have > **xw (the *-t- apparently > **d).  The usual explanation is "paradigmatic pressure" (numbers being a closed system), with sporadic assimilation of the **xw to the **f of following "*five"-- thus, Gothic has fidwo:r, fimf.  The reverse change affected Latin 4 and 5 quattuor quinque for quattuor **pinque or somesuch.  (IE '5' was *penkwe).  Sanskrit catur, panca show the expected reflexes, as do (I think) Gk. tessera, pente