Necropolis versus Orcs.

That's if the game is played as Deidra.

SA against orcs is only viable as Deidra, otherwise what's the point. An exception is having many castles and a mentor; then Lucia and hordes of princes could work, but only as an option.

Dragons will lose a lot if they charge at cyclops.

1. I don't care about the dragons; before 3.1 they weren't even built in the late game for financial reasons—it was easier to buy level 6 units from other castles.
2. And will they lose them? The retaliation is negated by ghosts with a 50% probability without punishment; there will be no retaliation with a probability of roughly 70%+, if the dragons' turn comes after the princes. Their initiative will be close, so they'll move consecutively. At worst, there are liches who are almost certainly blocked by that point, and therefore useless.
3. Dragons aren't saved; their job is to bite down all the key stacks.
4*. Need I remind you that banshees hit every stack adjacent to the bitten ones for 30 damage each, with no retaliation?

Well, in theory, you can create negative morale, but that's still no guarantee of victory.

-5 Luck/Morale on the orc's key stacks even at an early stage of the battle = victory for any race. Cyclopes move rarely, hit weakly, and their rage wears off quickly.

Goblins can be pushed a second time.

That is far from always the case. I'll use your own battle as an example: the chieftains moved before the hero, but the goblins had to be pushed. Standard layout: chieftains have higher initiative than the hero, while goblins' is comparable (the staff clearly worked to cut artifact bonuses to initiative). Please read the logic carefully and you'll understand what I mean.
By the time the Necromancer hero moves, it's already known whether the chieftains and goblins failed their morale check. Let's consider the options.
1. No one failed morale.
Then the hero has time to drop the mark with mass haste and move BEFORE even the chieftains, casting Subjugation on them. There's an even cooler option: if the initiative on the chieftains shows 11-12 (or less), you'll have time to cast Shout, mass haste, and Subjugation before their turn and the goblins' turn, obviously. Result: Subjugation lands in the first conditional round.
2. Morale check passed for chieftains, failed for goblins.
Immediately cast Shout. The hero's and chieftains' initiative will roughly equalize, but the chieftains will almost certainly move earlier next time due to the previously accumulated difference. There's no point in pushing the goblins; the mark is still working. Here, you either accept the scenario as in the point above, or press the Wait button and hope for RNG that you move before the hero again—fractional initiative values aren't displayed, so it's not exactly known. However, this won't delay Subjugation for long, since there will be no more positive morale.
Result: Subjugation lands in either the first or second round.
3. Morale check failed for chieftains, passed for goblins.
The hardest scenario to calculate. Shout comes first, but I can't say how many rounds before Subjugation lands. It might be the first if the gobs move earlier after morale than the hero does after Shout. Or it could be significantly later. However, the probability of such a layout is 1 in 4, and the probability of a truly bad outcome within that is roughly 1 in 2-4. Total: about 10% chance of problems with Subjugation.
4. Morale check passed for chieftains and passed for goblins. After Shout, the task boils down to point 1, just shifted by 0.5 rounds.

What could ruin all these reflections? The interference of a Barbarian hero; he can also push. There are more degrees of freedom, it's hard to calculate, so it's tedious. A rough prediction: it won't be easy for the Necro to drop the mark, which means it will take a long time.
But another question immediately arises: what will the cyclopes do without being pushed by the hero and chieftains, with their pathetic initiative, speed of 4, and morale/luck of -5? Correct: take a long time to walk and die quickly.

Yes, then my opponent didn't build up their Necromancer the best way, and I was only playing as orcs for the second or third time. So who knows how a battle between pros will turn out.

Everything we've been discussing here at length and tediously is something a pro would estimate in a couple of seconds and execute.

It seems it was the patch itself. I tried reinstalling the game and downloading the patch again, and now at expert Blood Rage absorbs 60%, as you said. No idea why it was different, most likely I downloaded some glitched patch that time.

Well, it's great that it's sorted out. But you understand that the damage hitting your orc through rage needs to be multiplied by (0.35/0.2)=1.75. That's a pretty significant correction that could have changed the outcome, or might not have.

Although there's another trick: you can collect a full dragon set and then get +1 7th lvl per day. In this regard, the orcs will have an advantage if both he and the necro collect it. The cyclops is far more powerful. And on that map, there was a good chance to collect it. I managed it as demons on the same map.

If the dragon set is collected, you place 2 stacks of shamans and 4 chieftains of 60 each. The hero is given tactics, speed aura, etc. Cyclopes tear everyone apart, regardless of race and the skill level of players on both sides. Fans of the mentioned "mad rush" tactic will weep with joy.
And only one question remains: why play a map where the outcome is known in advance and doesn't depend on the skill level? All races can be sorted by the strength of their 7th level unit from the start, and you don't even have to worry about it.

Summarizing the last ten posts in the thread:
When playing the late game in intermediate layouts (where the Necromancer has a meat advantage, but not a total one), "shaman wars" are the deciding factor.
The Necromancer must think in advance how to drop the mark and/or quickly wipe out the shamans, and the orc how to maintain the mark, pushing the cyclopes along the way, and protect the shamans.
Whoever handles their task better will very likely win.