End-of-line processing and attribute normalisation

I apologise for repeating an issue I already raised, but I see that
the proposed resolution of PE8 persists in the confusion.

Section 3.3.3 as amended in E24 says that the attribute normalisation
process takes place *after the end-of-line normalisation described in
section 2.11 has been performed*.  This means that #xD#xA has already
been replaced with #xA.  Why then does the attribute normalisation
process refer (in bullet point 3) to replacing such sequences with
#x20?

How can bullet point 3 ever apply if the end-of-line normalisation has
already been done?  Shouldn't bullet point 3 be removed completely?

-- Richard