Dear all,

Attached are the combined comments of Patrick, Marcel, Gerco and me.
Please, check if I changed you comments to a understandable version and 
is still what you meant.
I included 2 questions to Patrick.
Monday 18h I will upload our comments to CDS after my last check.

Best regards,
  Tjeerd

Congratulations for that nice new result. 
It shows the potential of conversions for such studies.

General: It is unclear where the uncertainties on the BF of the normalization 
modes enter the computation of the upper limits. 
Can we also give fitted yields in addition to the upper limits?

Method: 
Why do we use S/sqrt(S+B) in line 108, while we aim for a limit only? 
The Punzi method or something similar would have been much more appropriate.

We don't write in Eq.(1) how we treat the factor 2 of B0/B0bar with respect 
to B+. 

In the paper we present Bs->J/psi gamma and B0->J/psi gamma BRs.
The Bs are measured for the first time; the B0 result is roughly the same as the one
of BaBar.
Is this the reason why only the Bs result is shown in Fig.2 and in Fig.3 (and not B0)?
Probably the B0 result is too small to be seen in Fig.2, but we could show the 
CLs for the B0 measurement in a figure similar to Fig.3.

In detail:

Abstract:
"data corresponding to an integrated luminosity of 3 fb-1 collected by
the LHCb experiment" is short. Is it also agreed on to leave out the
pp collisions and the energies?

Reverse the order of the last three sentences about "B0s, B0, and B0s" to
"B0, B0s, and B0s".
(The order will then agree with the title. And the last sentence will be about
the same decay as the sentence before it.)  

Line 3: 
Replace "many observables require a good understanding" by
"the prediction/computation of many observables requires a good understanding"
(observables do not require anything).

Line 22: 
Please, put charges to both D's in "DD".

Line 25: 
On the front page it is said to be submitted to Phys. Rev. D. 
This is not a "Letter".

Line 31: 
Please, explain better how the converted photons "broaden the shape of background".
(You mean of course that of the reconstructed mass distribution.)

Line 33: 
Write "\B=\bd,\Bs decay vertex" here.

Line 34: 
Indicate which set of decays we use, 
avoid "fall within the same region" (jargon), 
and explain us why other B decays would not have the "same (reconstructed) 
mass region of the signal".

Lines 39, 203, 215, Fig.3:
"CLs" should be written with roman characters.

Line 61: 
Did you really only use Pythia 6?

Line 78: 
Replace "combining two opposite sign electron tracks, significantly displaced 
from the B vertex which can be associated to electromagnetic clusters in the ECAL" 
by "that combines electron and positron track pairs, which can be associated to
electromagnetic clusters in the ECAL and are significantly displaced from 
the B vertex."
(The tracks and not the vertex are associated to a cluster.)

Line 89: 
Replace square brackets [] in "m[jpsigamma]" by round brackets "()".
(In line 75 we also have pT(mu).)

Line 97: 
"The largest value of the BDT response for all tracks in the event is used 
as an input variable to the final selection" is not clear.
It seems that only one track is used in the final selection.
We also do not say if a large BDT value is a good or a bad value.
But Fig.4 later does help the reader.

Line 108: 
Replace "signal events" by "signal decays".
(Patrick, I like to remove this comment, because we do select events, not decays.
I do not understand the  assumption about the branching fraction here.
Therefore, I may misunderstand the whole meaning of this optimization.)

Line 112: 
Replace "K*0->K+/-pi-/+" by "K*0->K+pi-" and write the usual footnote
about "charge conjugation is included".

Line 120: 
Subscripts "sig" and "norm" should be written in roman characters.

Line 132: 
Replace "This effect is treated as" by
"This difference is taken as", if this is the case.
(Do not use "effect". These are statements. 
We don't see how we turn that into a systematic uncertainty.)

Line 134: 
Add "B mass" in "the B mass resolution".

Fig.2 legend: 
Replace "Bs->J/psigamma" by "B->J/psigamma".

Line 146: 
Add "+" to "B+ ->" .

Line 162: 
Add a "space" after "pi0".

Line 163: 
Please, indicate which kinematical variables you mean.
(Just "kinematics" is jargon.
 
Line 167: 
Add "uncertainty" in "systematic uncertainty".

Line 178: 
Replace "events" by "candidates" or "track candidates".
(This a very nice example of why "event" should be avoided here. 
All our "events" are "downstream". Patrick, what do you mean here?).

Equation (2):
The end bracket ")" of the second likelihood "L()" is misplaced.

Lines 222-224:
Correct typing errors "pseduo-", missing "," before "which", and "large". 

Line 216-217:
We suggest to replace this sentence by a Table with the 4 limit values.
(In both the Abstract and the Conclusion the same sentence is already written.)

Lines 223-225:
We mention explicitly that the channel is sensitive to right-handed currents 
and charged Higgs. Why are only these two mentioned? It could be anything, 
including left-handed W-primes.

Line 227: 
We should compare our Bd limit with with the one of Babar here.

Figure 3:
CLs is parameter which seems to be equal to CL-1 and unequal to CL.
Which is strange as often CL(s) can be found.
Maybe that a word of warning or a description of CLs can be given at the first 
reference to [6] in line 39.

Fig.4: 
Replace "MC" by "simulations" and
"MC events" by "simulated decays".

Line 253, 262:
Write the correct symbols as in the actual paper title of References [1,6]. 

Line 258:
Write "collaboration" in Reference [4].