This is also how the installation procedure was described in the pull request when the device was added
and someone mentioned the same procedure would be required for DIR-842 as well, though I couldn't reproduce it (not sure whether this is solved using the encrypted factory images, or maybe it occurs only upon first flashing to OpenWrt?).
Sorry for not updating this thread earlier, but I meanwhile got a DIR-859 for testing and my pull request to add encrypted images (that no longer need the quirky recovery method, which doesn't work with all browsers) for DIR-842 was updated to also include DIR-859, but as of now this is still open:
I actually intended to make a few more changes to DIR-859 (but only minor stuff e.g. spi-max-frequency and the switch reset gpio), and maybe also re-open the abandoned PR someone made for DIR-869 (high power variant in the orange/black case), which could share a dtsi with DIR-859, but I was waiting for the encryption PR to be merged first, rather than opening yet another PR for the same device, that would only cause a merge conflict when the other one would be merged...
The images for DIR-859 A1 and A3 should be the same however (here in Germany only A3 images are offered for download by D-Link), even if they made changes to the power amplifiers, this would be handled by the wireless driver based on the art partition (which seems to be detected correctly, otherwise you couldn't even set such high transmit power values), so this really must be driver issue, however I'm not really familiar with that.
Someone suggests it might be due to the value being calculated differently for devices with three antennas:
Could you compare the output of e.g. iw scan (if you're using Linux) for DD-WRT / stock vs. OpenWRT, to see if the capabilities are actually reported differently in the beacon frames?