i have implemented nearly all of the editorial comments exactly as suggested.
the following are the exceptions.  sorry that they are out of order...

top of page 6, "as for" does not make sense.  we used "in the same way as" instead.

figure 5.  the symbols are X's, not crosses.  since they are the only symbols on the figure,
the caption is unambiguous as it stands.  if you want to force us to make a change, please 
specify how the journal wants these symbols identified.  (exes? x's?)

page 21.  "shock" can stand as it is.

footnote 1.  the purpose of a footnote is to *not* interrupt the main text with nonessential
details.  no change.

table 1.  the timeseries are still available even if mode parameters are not.  we have right-justified 
the last two columns.  no other change.

first paragraph of sectino 4.1: the suggested phrasing is equivalent to the original.  no change.

3 comments at top of page 8:
"High frequency?": We agree that this was a bit carelessly formulated.
We were thinking of "medium-l" defined in the operational sense of the modes
for which the l-leaks can be resolved. We have simplified the text to clear
this up.

"No correlations?": It is a bit unclear to which sentence
this refers. Either way, correlations should not cause a trend with
frequency or any other parameters. Certainly the scatter could change with
frequency, but the trend should remain flat, as we say.

"Why?": We are not not entirely sure what the question is. Is
there a reason to assume that a fit, such as the one we perform, is biased?
We suppose that it could be, but surely it is a a reasonable expectation, as
we state.

Given this we believe that the statements are sufficiently accurate as
stated and that lengthy discussions of biases in ML estimators would not
benefit the manuscript. However, if we have misunderstood the comments,
please do let us know and we will see if we can improve the text.

page 10.  "error on the average" is common usage.  no change.

second paragraph of section 4.3: As we use a maximum-likelihood method, 
not Bayesian inference, we believe that the statement is accurate as it stands.

page 18. We have rephrased the text. As to why the widths are
different we do not know. This general lack of understanding of the root
cause of the various systematic errors is discussed elsewhere in the paper
and we do not see the need for a separate discussion here.