I am reading [_the PhD thesis thesis of Kenji Lefèvre-Hasegawa_](https://arxiv.org/pdf/math/0310337.pdf) and the correspongind [_errata_](https://webusers.imj-prg.fr/~bernhard.keller/lefevre/TheseFinale/corrainf.pdf) by Bernhard Keller, my question is about the first error found in the thesis. **Lemma 1.3.2.3 c** states 'the adjunction morphism $\psi:C\rightarrow B\Omega C$ is a filtered quasi-isomorphism' (C is endowed with the primitive filtration).  Along the way of the proof it is claimed that the
comultiplication on $\mathrm{Gr}C$ vanishes, and this is wrong.
If we denote $W=\mathrm{Gr}(S^{-1}C)$, the thesis claims

>$\mathrm{Gr}(B\Omega C)\xrightarrow{\backsim} B\Omega(\mathrm{Gr} C)$ is the sum of complexes $V_{i}=\displaystyle\bigoplus_{i_{1}+\cdots+i_{k}=i}{SW^{\otimes i_{1}}}\otimes\dots {SW^{\otimes i_{k}}}$

Are $\mathrm{Gr}(B\Omega C)$ and  $B\Omega(\mathrm{Gr} C)$ actually isomorphic?. If so, can anybody give some details of such isomorphism. I think it is consequence of the erroneous assupmtion on $\mathrm{Gr} C$, but I am not sure.

On the other hand, Keller's correction of the original proof uses a **decreasing** filtration on $B\Omega C$ named $F^{'}_{l}$, by describing

> $F^{'}_{l}$ is generated by all $m$th tensor powers of $C$ where $m\geq l$

I find the description of $F^{'}_{l}$ is clear enough but I do not know how decreasing filtrations work (may they induce isomorphism in homology?). I would really appreciate some clues about these inquiries and a short clarification of what is either rescued or discarded from the original proof, in the corrected proof.