I'd nominate most of new-toy-syndrome 'Mechs.
Like, the Artemis IV upgrade for the Whitworth.
Or the Goliath upgrade with two tons of MG ammo. "Prototype rushed to production" works only for so far.
There are good ones around to be sure, but so many of them don't make sense neither out of universe, nor in-universe by any reasonable metric. This leaves irrational reasons, so figuring them out is basically impossible.
The Hellbringer comes to mind as well, though i suppose it can be justified by the fact that without endo-steel or ferro-fibrous, it is relatively easy to repair, and its nature allows for quick duels ("time-economical"), well suited for the Clans. But given that the Clans are perfectly capable of producing designs that are smart and effective, it is really baffling. Then again, i seem to recall that the design competed against even worse designs before it got selected for production...
The Hercules isn't too bad design overall, with its speed and primary armament, but its secondary weapons array is somewhat baffling. Small pulses in the rear? Yes, they're accurate but for anti-infantry use it would've made sense to locate them in the left arm, that way twisting torso would allow shooting more or less behind the 'Mech to keep infantry out of your back. The singular Streak launcher is also a bit dubious, something the field refit partially corrects by adding another Streak by removing the AMS and its ammo.
I'd argue there are some in-universe factors that we don't really see on the tabletop.
While it's easy for us to max out the armor on a custom design, I don't think it's supposed to be that simple for the engineers in the game world. While a 65 ton mech can carry a max of 13 tons of armor, I like to think that means squeezing every last possible bit of protection onto the frame. As in, you're probably making compromises in the design elsewhere.
If you think of a mech not just as a list of numbers, and instead as a complicated combat vehicle with thousands of moving parts, it makes more sense. I used to write out my mech designs on a piece of paper, and it looked something like this:
65T
-13.5T 260 engine
51.5T
-6.5T internal structure
45T
-6T gyro/cockpit
39T
-12T armor (192)
27T
And so on down the page. Like that, optimization is easy. But if you think about building it as in drawing up the technical diagrams, and figuring out where ammo feeds are going to be, etc, it suddenly gets a lot more complicated. While a particular design might be perfectly legal on the tabletop, that doesn't mean it's going to be easy to make it work in the universe. The Bushwhacker had notorious problems with engine and weapon placement, even though it was a standard 3025 design. I'd argue that certain under-armored mechs may have been designed with the intention of carrying more protection, but once the chassis was finished they found out that the placement of certain components prevented them from up-armoring the thing like they wanted. This sort of thing happens with real life military equipment, and you basically find out when it's too late to fix it.
And maybe the production model of the Hellbringer is just a nice budget heavy Omnimech. It's cheap to manufacture, fast enough to keep up with better designs, and carries a big punch. It uses a lot of the same parts as the Summoner, so you can keep your costs down even though it's relatively easy to damage. The logistical benefits of the mech outweigh the weaknesses in the design, so it's not really worth changing.
With the Hercules, there are a few mechs (like the Falcon) that put anti-infantry weapons in the rear. So there's some history there. Perhaps a favored tactic of anti-mech infantry is to stay hidden and then attack from behind? While the game rules allow you to torso-twist, it's possible that it's still not quite as effective as dedicated rear-mounted weaponry.
