Results on Linux:

With Python 2.7 trunk:
>>> print(', '.join('%s:%s' % (c, c.isalpha()) for c in u'абвгдеч'))
а:True, б:True, в:True, г:True, д:True, е:True, ч:True

With Python 2.5.1:
>>> print(', '.join('%s:%s' % (c, c.isalpha()) for c in u'абвгдеч'))
а:True, б:True, в:True, г:True, д:True, е:True, ч:True

With Python 3.0 trunk:
>>> print(', '.join('%s:%s' % (c, c.isalpha()) for c in 'абвгдеч'))
а:True, б:True, в:True, г:True, д:True, е:True, ч:True

Are you sure that you really typed the character "ч"? Can you retry 
using unichr(0x447).isalpha()?

Test with Python3:
>>> print(' - '.join((r"\u%04x" % x) for x in range(0x400, 0x4ff+1) if 
not chr(x).isalpha()))
\u0482 - \u0483 - \u0484 - \u0485 - \u0486 - \u0487 - \u0488 - \u0489

Which means that Python thinks that all unicode character in range 
U+0400..U+04ff are letters except the range U+0482..U+0489 (thousands 
sign ҂ to million sign ҉).