According to the Hanahan transformation protocol (one of the highest competencies published), a transformation of 200µl will contain $1×10^8 - 1.7×10^8$ cells.
Assuming that the pBR322 plasmid is used, 1ng of plasmid will convert to 0.36fmol of plasmid, or $6×10^8$ plasmid copies.
As we can clearly see, even at very low quantities of DNA used, there are multiple plasmids per cell. Since transformation is thought to be mediated by DNA channels, massively increasing the quantities of DNA would not equally increase transformation efficiency once all of these channels are saturated.
The example from NEB Turbo DH5α cells seems to show this quite well, that once DNA quantities exceed 1ng, the transformation efficiency drops precipitously.
This being said, if the amount of plasmids available is very large, there is no reason to avoid adding more than 1ng of DNA, as they would not reduce the total number of transformation colonies, but only the efficiency.