I think you misunderstand quite a bit of what dailytech really said and that is basically the issue. You compare different XXnm to each other by dividing the numbers but they really are more like names for a current technology product.
Here is a list of how the different process compare to each other in two feature size metrics.
If you read that you will notice that comparing 32nm Intel and 28nm TSMC is quite problematic. Then there are even more issues because Design guidelines ultimately also influence how closely you can pack logic together. A 14nm Intel Broadwell CPU has 1.3B transistors, the 20nm A8 about 2Bm which seems that Intel cannot do all that dense chips either. I.e. if you have to double or triple this or that via/gate whatever little feature to get a decent yield, you end up with less density even if your very smallest gates are smallest. Yes they are different types of SoCs, but both have a CPU, GPU and the remainder is similarly dense stuff, like media and image processors.
TSMC claims its processes are far better than Intel marketing makes them out to be.
One way or another the term 14nm should be considered as a name like Core i7 and is very loosely correlated with any real metric that can be easily compared to each other. It is just misleading math that isn't worth much. You should also notice that the numbers you picked from the article don't match the calculations you did with the rest.
Ulitmately you are right with the gist of your post just the math part I had an issue with.