(i) 14 Mar → 13 Apr: 30 days; 13 Apr → 12 May: 29; 12 May → 11 Jun: 30; 11 Jun → 12 Jul: 31; 12 Jul → 9 Aug: 28.
(ii) 12 July. The first three gaps are 29 or 30 days, but the two gaps on either side of 12 July are 31 and 28 days, breaking the pattern. The note says it was guessed on a cloudy week, not observed. From 11 June to 9 August is 59 days = 29 + 30, so the real purnima was most likely on 10 or 11 July.
(iii) Adding 29 or 30 days to 9 August gives 7 or 8 September. One month alone could mislead: a single gap of 28 or 31 days (like those around the guessed entry) would give a wrong idea. Many months of careful observations show the true pattern of 29 to 30 days and let us spot a mistake. The book says that "by combining careful observations of sunrises, sunsets, and lunar cycles, various calendars came into being".