The "additive effect" is something of a red herring. In fact, the dilution effect is stronger. Think about it this way:
There are probably dozens of genes that create the individual human face. A page back we were talking about COL17A1, which has been shown to affect distance between the eyes.
Henry Tudor had two copies of that gene. One or both of them gave him the specific distance between his eyes. Now, suppose that someone has 6 or 7 descents from Henry, and that those descents are all maybe 15 generations ago.
At each generation since Henry there is a 50/50 chance that a child will inherit the gene for that specific distance between the eyes.
Someone pointed out a few weeks ago on another thread that you can model this for yourself. Flip a coin for each generation. Heads, the child got the gene. Tails, the child didn't get the gene. Do this 15 times, once for each generation. What are the chances you'll get heads 15 times in a row? That's the basic probability of getting that gene with one line of descent.
Now repeat this for each different line from Henry. If there are 6 or 7 lines from Henry, flip the coin as many times as it takes to get a result for each generation along every line. What are the chances that you'll get heads 15 times in a row on one of them? That's the "additive affect" you get from multiple descents.
Now do the same thing for a dozen different facial features. What are the odds that you get heads 15 times in a row for all 12 features? That's the odds of ending up with Henry's face through one of descent. So, do it again for 6 or 7 lines of descent. The odds are better, but the odds are still astronomical that you'll get heads 15 times in a row for 12 different features, even if you repeat it 6 or 7 times for each line of descent.
Of course, the problem is complicated by the fact that Henry Tudor's version of the COL17A1 gene wasn't just his. It would be a common variation, one of an unknown number of variations floating around in the general population. So, there is also a chance that someone in all those 15 generations will get the same variation, or something similar, from a non-Tudor source.
If that happens, it might look like proof that the Tudor COL17A1 survived to a recent generation, or that it was added back in from a different Tudor line, but really it's from another source. And, because it doesn't have to be the same gene, just something similar, the odds are actually better that the Tudor face can be assembled out of the general population from non-Tudor sources than that it can survive the generations using just Tudor genes.