While you're correct that in 2016 the pattern was weaker (a >10% drop was present but only for a short time, about 2-3 weeks, so you could say it did not leave the normal fluctuation range),

427,392   2016-08-29 17:11:01   220,755,908,330 - 220.76 G   + 1.56 %   0x1804fb08   09 min 51 s   1.58 EH/s
425,376   2016-08-15 21:59:14   217,375,482,757 - 217.38 G   + 7.67 %   0x18050edc   09 min 17 s   1.56 EH/s
423,360   2016-08-02 21:50:42   201,893,210,853 - 201.89 G   - 5.43 %   0x18057228   10 min 35 s   1.44 EH/s
421,344   2016-07-19 02:22:42   213,492,501,107 - 213.49 G   + 0.04 %   0x18052669   10 min 00 s   1.53 EH/s
Halving Date: July 9, 2016
419,328   2016-07-05 02:16:01   213,398,925,331 - 213.40 G   + 1.88 %   0x180526fd   09 min 49 s   1.53 EH/s
417,312   2016-06-21 08:18:58   209,453,158,595 - 209.45 G   + 6.83 %   0x18053fd6   09 min 22 s   1.50 EH/s
415,296   2016-06-08 05:41:58   196,061,423,939 - 196.06 G   - 1.63 %   0x18059ba0   10 min 10 s   1.40 EH/s

btw: Is your quote related to the 2020 hashrate evolution? If you mean the 15% drop was earlier (or later?), there always can be some delays, in either direction.

What I can't explain at this moment is the sharp November drop (this was ATH time in '20 ...).

495,936   2017-11-24 21:53:16   1,347,001,430,558 - 1.35 T   - 1.28 %   0x1800d0f6   10 min 08 s   9.63 EH/s
493,920   2017-11-10 17:13:51   1,364,422,081,125 - 1.36 T   - 6.09 %   0x1800ce4b   10 min 40 s   9.75 EH/s
~
552,384   2018-12-03 13:59:28   5,646,403,851,534 - 5.65 T   - 15.13 %   0x1731d97c   11 min 47 s   40.40 EH/s
550,368   2018-11-17 01:51:24   6,653,303,141,405 - 6.65 T   - 7.39 %   0x172a4e2f   10 min 48 s   47.61 EH/s
~
604,800   2019-11-21 19:08:52   12,973,235,968,799 - 12.97 T   + 1.99 %   0x1715b23e   09 min 49 s   92.81 EH/s
602,784   2019-11-08 01:30:36   12,720,005,267,390 - 12.72 T   - 7.10 %   0x171620d1   10 min 46 s   91.02 EH/s

But eventually a new equilibrium will be found. Let's see how it plays out.