This is a long standing bug - in this instance, the addon for some reason thought that he has some legendary boots equipped, which is untrue. It happens virtually every raid, that the comparings are wrong and therefore rankings are unuseable. This reset, already three times - two times, a random legendary was chosen instead of one that is actually used (it compared a belt to the resto druid legendary belt, which I haven’t equipped in months), then this boots, and once it was some thought-up non-legendary.
This time the GEAR optimizer uses random legendry to “optimize” gear, while actually making the gear worse.
To illustrate: 910 (+61) relic Grovewalker (+all HoT healing) is considered worse, than 915 (+62) Armor of the Ancients (+Ironbark damage reductuion). That is rather obviously and clearly wrong.
When will those bugs be fixed, if at all? No response in a week… 
Another ilustration about how much the gear optimizer is wrong: only 2 of the 74 top M ToS log use the belt. That’s 2.7%. ( https://www.warcraftlogs.com/rankings/13#metric=hps&class=Druid&spec=Restoration )
For H ToS ( https://www.warcraftlogs.com/rankings/13#metric=hps&class=Druid&spec=Restoration&difficulty=4 ) - 1 out of 90, that’s 1.1%.
Getting to a round of changes to the team optimizer is on the list, but we have a few other things stacked up.
The main thing is our new “version 3” gearing strategies, which will be much better at choosing legendary items and adapting to your chosen talents.
As for “comparing to random gear”, I can’t reproduce this. I would need steps to reproduce in order to debug it further.
Usually this happens when some information is out of date for a character… the team optimizer is currently quite sensitive to every player on your team keeping things very up to date. That is one of the things we want to think about for the next version of the team optimizer – making it a bit easier to use.
For now, most problems can be fixed by ensuring that every person has the addon, every person has every spec updated and configured as desired, and every person has a well-maintained public profile. It really is a tool designed for an entire team that wants to do data-driven optimization.
An account that had up-to-date stat weights suffered from having random legendaries compared to.
The screenshot is proof enough.
IF it is, it malfunctions terribly. Let me give you an example: car racing. A team asks “Which car will do the job the best?” First thought, a racing car. Common sense thought, a racing car. Simple spreadsheet, a racing car. Proper simulation, a racing car.
But YOU claim to use advanced simulations… and suggest: racing horse.
The tool is a misleading burden even to people who have it “updated and configured as desired” and “well-maintained”.
Like a compass that is not pointing north, but west.