Which software are you using to calculate this?

These calculations were done the hard way, with a TI-82 and combinatronics formulas.

Could you add the weighted mean deviations for cards and dice?

I can add the deviations for dice, but I am still figuring out the distribution curves and deviations for the cards.

Would it be possible for you to calculate the same stuff for when two and three players are drawing from the same deck? Do you think that the average result  would increase, decrease or remain the same with a higher number of players?

While these calculations are possible, I don't think they would be useful. With multiple players drawing, for example, each of the 63 possible outcomes for the first player would generate 63 possible outcomes for the second, meaning that there would be 3969 probabilities to calculate for player 2. The average result for player 2 would depend entirely on the draw for player 1.

I really like your proposed rule. Would the Raise obtained with "numerical card +  Arcana" count double when is possible to obtain two Raises with a 15?

I don't think it would, only because the odds of getting a total of 15 or above on two cards are significantly lower than getting a 10 or above. I would consider allowing it with any two numerical cards, however, as the odds of getting 15 or above on three cards are fairly good.

How much the mean would increase if you could get Raises by combining two Minor Arcana cards?

I'll crunch these numbers shortly. The issue is that the number of Raises from a given hand (or roll) is not entirely dependent on the total result of the roll (e.g. a roll of 9,9,9,9 totals 36, but only yields two Raises, while a roll of 10,10,8,8 has the same total but yields three Raises). There is not, so far, a good way to model Raises. On the whole, however, you average a little under one Raise per two cards (based on some rudimentary calculations) so I expect that the mean number of Raises would increase measurably if you allowed two Arcana to be combined for a Raise.

There are special effects that add or may add Raises to the total in particular situations (Dame of Cups, Queen of Coins and King of Coins). Given that the cards can be used when drawn, how much would those effects affect the mean values?

Only two of the 16 Minor Arcana directly grant Raises, the Dame of Cups and the Queen of Swords, and while the odds of getting either change based on the number of face cards you draw, you have a 12.5% chance to draw one of the two on your first face card. The problem here is "given that the cards can be used when drawn:" the Queen of Swords (among others) specifies that it must be played before a Risk, meaning it cannot be played for the Risk it is drawn on. This leaves the Dame of Cups, which can only be played " after you make Raises for a romance-themed Risk;" calculating how often that happens would be needed to give a meaningful statement about how much the card affects mean values.

I know it is hard to say, as it depends a lot on the players, but it would be interesting to see how the mean changes when the players keep high value cards (let's say 8-10) when they hand is higher than the mean and use them when the hand is lower.

I think Hero Point economy would limit that behavior substantially. A more interesting question, I think, would be how often players keep Minor Arcana.