Dear Julian, dear proponents,

Thank you for the many additional cross-checks performed.
We would like to respond to some of your answers below:

>> Section 1: Introduction
>> Line 11-13: Although you mention CP violation in mixing as a possible source for a non-zero cosine term, you never
>> introduce the accompanying parameter q/p, whose deviation from unity measures this directly. We therefore suggest to 
>> leave out this remark.
> We do not understand this comment. With the same argument we could as well not mention the other source of a non-zero
> cosine term, the direct CPV, as we do not introduce |A_f|/|A_fb|. The same would hold for not talking about interference 
> of mixing and decay because we do not introduce lambda_f.
The main difference between both sources:
CP violation in B-Bbar mixing is usually measured with semi-leptonic decays and so far assumed negligible in all B decays.
Hence, one would not immediately explain a non-zero value of C as due to B-Bbar mixing.
On the other hand, we know that contributions from penguin topologies must enter at some point and a first signal of them
would be a non-zero value of C.
I hope that motivates our initial comment a bit better.
We are fine with whatever you decide: leaving the sentence as is or rephrasing it.

>> Section 2: Data samples and selection requirements
>> (1) In case multiple candidates in a single event survive your selection cuts you keep only the one with the smallest 
>> chi2 of the decay time fit. Does this not introduce a bias in your data sample? After all, it is possible to have an 
>> event A with two candidates whose chi2 are both smaller than that of a single candidate in an event B. If this does not 
>> introduce a bias, please explain why this is effectively a random choice. We suggest to follow the guidelines written 
>> down in the internal note LHCb-INT-2011-009 entitled "Dealing with multiple candidates".
> The choice of a best candidate selection based on the chi2 of the DTF is a result of closely following the selection in 
> the phi_s analysis. The best candidate selection removes <4% of the candidates. For a cross-check, we have prepared a 
> tuple with a worst candidate selection. Assuming that 96% of the sample consists of the same candidates as before, we 
> calculate a deviation of the fitted parameter values of
> Delta S = 0.012 ± 0.019
> Delta C = -0.028 ± 0.026
> These numbers are in good agreement with no change. Hence, our best cand. selection seems to be effectively random.
Ok. We think this should be added to the systematics (even though the effect is small) for completeness.

>> Section 6: Systematic uncertainties
>> (1) Can you motivate why comparing with an sFit is an appropriate way to extract systematic uncertainties due to 
>> background decay time modelling?
> Do you mean in the paper, for example by referencing the sFit paper by Yuehong?
Our answer?

>> (2) Your background model systematic, as extracted from difference between a standard fit procedure and the sFit 
>> procedure, seems very large. Is it possible you are overestimating this systematic? You have a very small number of 
>> background events compared to the signal, and almost nowhere in the considered mass range are you signal free. In this 
>> case we expect the sWeight procedure to be quite poor, and thus show a larger statistical dependence on these few 
>> background events. It could therefore be that the difference between the two methods is not representative for a 
>> systematic bias at all, unless the central values disagree.
> Right. The "sFit"/"cFit" comparison feels like we are overestimating the systematic uncertainty. However, we can not 
> think of a better way to estimate this syst. uncertainty. We are of the impression that this is a more valid approach 
> than arbitrarily choosing different parameterizations and comparing the results. Additionally, as the derived syst. 
> uncertainty is not the leading syst. uncertainty on S and C is clearly statistically limited we think that this is not a 
> big issue for the final result of this analysis.
Our answer?

>> Line 65: What does "clearly associated to one PV" mean?
> To clarify this selection requirement we would have to add the max ip chi^2 values for the best PV and the min ip chi^2 
> for the next best PV. We doubt that anybody outside LHCb cares how we actually do this.
We disagree; these selection cuts have already been discussed by other LHCb papers, so we see no reason why you cannot add
them here too. See for example LHCb-PAPER-2012-019, section 3


We are happy with all other replies.

Kind regards,
Kristof