A problem with PIE *p(o)lH-

Gr. _po'lis_ 'town' is routinely compared with Lith. _pili`s_
'castle' and Skt. _pu:'r_ 'fortified place', and on this base
something like PIE *p(o)lH- 'fortified place' is usually
reconstructed. I have no problem with the Old Indo-Aryan and
Lithuanian lexemes (the last even seems to point to a possible
etymology -- cf. Lith. _pi`lti_ 'to pour, to fill' (< *plh1-) and a
deverbative _py'limas_ 'rampart', lit. 'what is piled up' (here -yl-
from -il- by Schleicher's pseudo-apophony, -yR- < -iR- (R - sonorant)
being one of the most trivial cases); the Greek one, however, has an
alternative form -- _pto'lis_ 'id.', and if Myc. Gr. _po-ta-ri-jo_
does belong here, _pto'lis_ seems to be the older one, making a
comparison with the Old Indo-Aryan and Lithuanian lexemes at least
problematic.

Any ideas?

Sergei