Likely the REAL reason the Radeon Pro 560 sucks in the MBP 15" of mid-2017 is because of thermal throttling.
The equivalent 2016 MacBook Pro uses a Radeon Pro 460 (which I have). It has 16 CUs (1024 shaders) running at 850 MHz stock* with an 80 GB/s memory bandwidth to the 4GB of GDDR5. By the way, the TDP is 35 watts. It thermal throttles in most games that mostly max both CPU and GPU, especially in Windows.
The Radeon Pro 560 in the mid-2017 MBP, by comparison, has the same 16 CUs (1024 shaders), but running at 925 MHz with a slightly improved 81 GB/s memory bandwidth to the same 4GB of GDDR5. It also has the same TDP of 35W.
Now, I suspect that the small performance drops are due to throttling because since the two MBPs are identical except for a few faster components in the 2017 model, the hypothesis is that the second one throttles more. Why do I think this? Well, the core clock is increased 75 MHz with no known increase in cooling efficiency. Thus, the 560 will run at higher clocks more of the time, causing heat to go up, and eventually, throttling to ensue quicker. The benchmarks take a non-zero time to complete; thus, the 560 likely has to lower its clock for more of the test, resulting in a slightly worse score.
*Under Windows, using experimental modified drivers, overclocking IS possible on both models (to some extent). When using MSI Afterburner, I was able to get a stable overclock by increasing the 460's clock to 560 level, dragging the memory clock slider all the way to the right (!), and undervolting by 40 mV. The card throttles about the same amount, but performances is up on the 560 by about 5%. Alas, it doesn't persist through reboots, but that's OK for me.