Trying to figure out how Assault works

[Xerxes77]

So after getting a series of wild results when attacking enemy positions, I decided to look into the actual numbers and test a few standard situations. For that I created a simple Grenada scenario with matched units and let them have at each other in a series of test assaults.

So far my findings have been... weird, and I wonder what I'm missing. This is a bit long, but I'm really intrigued.

Essentially, losses resulting from assaults seem to be all over the place. I understand the results are intended to be swingy and unpredictable to a point, as they're based on a number internally selected at random from a wide (but known) range of values. The problem is I'm getting casualty rates that are well outside that range.

Example 1:


In this test scenario, my 10 American soldiers have just assaulted an enemy unit of equal strength and successfully dislodged it from its position. The outcome was 9 to nil, so I don't think they're popping the champagne just yet. Still, that went better than...

Example 2:


Another successful assault, at the cost of a complete squad wipe-out.

Here are my main questions:

1) According to the manual, defending units should never retreat UNLESS they've suffered losses, after which morale calculations come into play. See Example 1 for a counterpoint. I've seen several cases where this is simply not the case and they will retreat at full strength.

2) If the assault resolution works as outlined in the manual, as far as I can tell the attacking unit should simply not suffer that many losses. The process is a bit involved but clear enough: the infantry assault value and the weapon assault values of the defending unit are multiplied by their respective strength and normalized against a baseline (1/10000 for the whole SB series) before being applied to the Attack Low/High Loss parameter values (24-240). This establishes a range of possible losses for the attacker. Unless there's some other modifier involved not mentioned in the manual, I cannot explain the relatively common occurrence of casualties exceeding the higher endpoint of that theoretical range.

3) Perhaps more worryingly, assaults against Pinned and even Demoralized units did not seem to yield significant differences in my tests. According to the manual, pinned units defend at 1/4 value, and demoralized units at 1/8, so the difference in maximum losses should be visible. Yet in the relatively small set of tests I did, pinned and demoralized troops seemed to hold their own in this regard (not so much for retreats, which are based on morale). 2 pinned guys were perfectly capable of killing 9 healthy attackers, which seems to go against the base assault resolution math.

I know there's an optional rule that flattens this volatility a bit, but I'm trying to grok the basic assault mechanics and I wonder if there's some giant obvious factor I'm overlooking?