I finally had a chance to dig into this and see what is going on. I’m going to discuss only the profit per policy but the arguments apply also to the loss ratio.
How are average KPIs computed? 
The average profit per policy is computed in the following way:
This computation is a representation of your “average market”, not your over all performance. Now what is happening in your case is that while in most of your markets your profit is positive (as seen by both the average profit per policy and your KDE plot), when you do lose money, you lose a lot. Hence your average profit is negative over all markets while your average profit per policy is positive.
Put in another way, you can see that when your profits are negative, your marketshare is also larger, hence your negative profit per policy values have a larger “weight” in your final average profit.
Why not a simple average? 
You might be wondering then, why don’t we just give you the cumulative average such that the average profit per policy would be weighted by how many contracts you win in each market.
The reason we don’t do this is because this discrepancy is informative. It is telling you that “most” of the time you are doing fine but when you lose money, you lose (relatively) a lot.
I hope this clears things up a bit more and thanks for bringing it to our attention! 