The following were the raw marks:
arr := 
[44, 43, 40, 39, 38, 38, 38, 37, 37, 36, 35, 33, 31, 29, 29, 27, 27, 27, 15];
In my report, I recommended the scalings:
 35 --> 70%,
 27 --> 60%,
 15 --> 50%.
 This would give 11 first class marks, 7 2-1 marks and 1 2-2 mark.
My feeling is that all but one student had a script of at least 2-1
standard (which is also consistent with the fact that almost all
students are of at least 2-1 standard in order to be allowed to do Part C).
If they come back to me with a proposal for a different first class
line, I'm ok that the scripts on 35,36,37,37 are borderline,
but the three students on 38 were very clearly first class standard
(and really the two on 37, it would be good to maintain them as first class).

The following were the raw marks multiplied by 2 (to give raw percentages):
arr2;
[88, 86, 80, 78, 76, 76, 76, 74, 74, 72, 70, 66, 62, 58, 58, 54, 54, 54, 30];

  My report was as follows.
----------
Section A. Confidential Report for Examiners.
  Everything went well. There were no changes to the marking scheme.
  - Raw mark which might be scaled to 70: 35.
  - Raw mark which might be scaled to 60: 27.
  - Raw mark which might be scaled to 50: 15.
Section B. Comments on the paper for including in the Examiners' Report.
  There were 19 candidates; 17 answered Q1, 12 answered Q2
and 9 answered Q3. Q1 was well answered. In Q2, few candidates
spotted the points of order 4 in Q2(b), and in Q2(d) most candidates
got the upper bound on |y| but many had trouble with the upper
bound on |x|. In Q3, parts (a),(b) were well answered, but many
candidates found part (c) difficult.
----------