Regarding the Letos vs. Axel matchup – if the player using Letos had used a teleport spell to the Witches at the end, they would have won easily. And they should have used Vampirism. Axel has the ability to cast Remove Magic 3 times + Adepts can cast it once more, and he should actively use this (by actively, I mean not mass removal or removing Plague from every stack, but only removing Plague from Paladins and removing some enchantments – save mana for the end of the battle when Letos has little mana left; he always casts Vampirism or Plague, so that's what needs to be removed). Also, in that battle, the Knights should have removed magic from the Angels so that they would have lasted longer. There is a chance that Letos simply runs out of mana, and then it's just a matter of technique. If there is almost no Plague in the first turn, you can try casting Mass Haste and hope for the Paladins' morale.
I watched the Carle vs. Orson battle, and in my opinion, Orson shouldn't move anywhere at all, but just stand still and kill all of Carle's troops with direct attacks from the hero or with crystals (kill the War Wagon and the Ballista). If Orson's troops start to move forward, they will most likely die from Direct Strike. And we should kill Spear Masters first, not Harpooners, although Harpooners hit harder from a distance, but they have fewer shots, and the Masters are much stronger in melee combat later on.
Regarding splitting the Griffins, this tactic is not very effective in the Axel vs. Letos battle, because small units are easily destroyed there. Also, if you cast Vampirism on them and attack with spells, it will be much more effective. 1 Witch casts Dispel on the Archers, which greatly reduces their effectiveness; you either have to spend Remove Magic on this, and there are only 4 of them, or waste a turn of the Paladins, which is a pity. But there are effective tactics with splitting the Griffins, for example, as strange as it may seem, this tactic is quite good in a battle against Iris, especially when combined with Slow. Because the Griffons of Iris will most likely dive on Archers or Witches at the beginning, and not on the split and relatively safe stack, so the probability of guessing where they will dive increases. Then they reach and block the Archers.