Pelton, take a deep breath.... You may be right there is some problem with the algorithm, but you are getting into hyperbole territory again and that is when you are less useful!
It would be great if you could run these attacks more than once, but I guess it is a server game so no? Just to get a sense of where these attacks lay on the spectrum, because you assert they are normal, but players have a strong tendency (not just you) to remember 'abnormal' results and thus make them feel 'normal'.
The other problem is, I am not sure these are bad results from a military operations expectation perspective. In the first one, a massively superior force, in every way, conducts an assault against hasty positions, exploits the initial breakthrough with its huge armor advantage, and creates some havoc with the withdrawing defenders. It still lost 2:1 in AFVs, which is probably appropriate for an attack. remember that artillery in the game includes short range mortars and AT guns, not surprising a large armored assault gets good kills on those.
The second is a very large infantry force attacking a much more mobile defender. Looking at just "man" losses, you see the attacker suffered 4% losses, while the defender suffered 6%. but the problem with that analysis is that the defender really isn't made of "man" type defenders in the main, it is a highly moblile force, tanks and cav. here we look at AFV losses and we see that the defender took 22% losses. That is actually a very good result, especially when you consider that they are being attacked by a completely infantry force. From a realism perspective, it is not really so off, you could expect to see losses as the slower attacking force closed on defensive positons, but there would be almost no overrun or pursuit losses, which are the big killers in this game design.
So, I get the issue you are bringing up, but I am not sold that it is as completely broken as to call it 'middle earth', which seems to be your current favorite 'it is broken' tag.