Of course I'm stuck on the "more cards" concept as that is what any rational person would do. Of course a 6-card 1080 rig will beat a 6-card 1070 rig, because 1080s are faster. Even a 5 year old can figure that out. If that is all you are trying to convice me of, then don't bother. I agree 100%.
So let's explore the dumb idea of taking the cost of a 6-card 1080 rig and instead using the same money to build 1070 rigs (No one has an unlimited budget and if you did, why bother to mine? Even if the budget is much greater all the numbers here will just scale proportionally). Let's assume that the cost of the rest of the rig equals the cost of a 1070...or $400 USD:
$600 x 6 + $400 = $4000 (The cost of a 6-card 1080 rig)
Now, being the neanderthal that I am, I decide to take that same amount of money and buy 1070s instead, because "Oooh, I can has even more shinies!"
So I spend $800 on two barebones rigs and have $3200 left over and buy 8x 1070 cards. Well I must be dumb, because I have 33% more cards but they get 33% less PPD, but I'm also paying more for power!
Ignoring the electric difference (~$15.77 USD per year for *total system* @ $0.01 per kWh):
GTX 1070
Cost: ~$400 USD
PPD: 600,000
FLDC per month: 12,000 - ~$14.72 USD
CURE per month: 787 - ~$35.43 USD
Purchase ROI: ~8 months "raw cost" per card
Yearly Revenue: ~$599 USD
GTX 1080
Cost: ~$600 USD
PPD: 800,000
FLDC per month: 16,000 - ~$19.29 USD
CURE per month: 1050 - ~$47.25 USD
Purchase ROI: ~9 months "raw cost" per card
Yearly Revenue: ~$798 USD
The only way 1070 tops 1080 is with a constrained initial "budget" where one would be buying 6x1070 vs 4x1080; however, in that narrow of a constraint, ROI isn't an actual concern (as it's trumped by initial purchase limitations).
*total system* revenue of 6 cards is 1070 ~$3,593.48 per year and 1080 ~$4,791.31 per year.
So let's use your own numbers AND give the 1080s the benefit of the doubt and assume they draw the same amount of power per card as a 1070 rather than 20% more. I will be paying $21.03 whereas you are paying $15.77 (something tells me this number is wrong, but of course you did the math in every way shape and form), yet we are both making $4791.31 per year as you say.
Now we can stop here, and again it's easy to see that the 1080 system comes out ahead by almost $6/year. Wowza!
But, for some idiotic reason I decide I would rather have more income, so once I earn enough to add more cards I do that. At $4791.31 per year, I will earn that amount divided by 365.25 per day, or $13.12...minus the cost of power ($21.03 / 365.25 = 5.7¢), so $13.06. After 30.6 days I will have enough to buy another 1070. Let's call it 31 days to simplify things and err on the side of the 1080 system. So one month in I am now earning [($4791.31 - $21.03) / 365.25] * 9/8 , or $14.69 per day with the power cost now factored in, whereas you are still earning ($4791.31 - $15.77) / 365.25 = 13.07 per day and have accumulated $405.32 in profits, and (thinking maybe there is something to this "more income" shenannigans) buy yourself a shiny new rig but which is devoid of GPUs (or don't. Whatever).
Shall I continue?
After 27.22 days (let's call it 28 just 'cuz) I buy myself another 1070, and am now earning $14.69 * 10/9...or $16.33...per day, and you will have accumulated $771.13 (or $13.07 x 59 days) and spent $400, leaving you with $377.13 and still only 6 1080s and an income of $13.07....
Do I really need to continue at this point? I can but I'll stop here. Clearly you know how this ends up as you have already done the math in every which way. I'm glad you don't have buyer's remorse and are happy with your purchases, but to imply that anyone that does different (and you DID imply this) is somehow not smart, is simply wrong. Good day!