In p. 18, it says: "consult the appropriate column, determined by the knight’s rank and pre-existing male heirs."

However, the explanation on how the pre-existing male heirs affect this table was left on the cutting room floor.

What you should do is to shift one column to the left if the knight already has male heirs from a previous marriage. So a vassal knight with heirs would roll from the household knight column, and so forth.

Also, in those rows where you have an eldest daughter of X or a younger daughter of Y, you should, generally, use the younger daughter of Y in those cases.

The reason for both of these is simple:
The eldest daughter gets the biggest dowry, so the fathers are looking for the wealthiest man they can set her up with, and naturally they want to ensure that their grandson will be the heir, which will not happen if there is already a male heir around from a previous marriage (girls don't matter). Hence, most of the 1st marriages of vassal knights with be with the eldest daughters of their peers, and the second marriages tend to be with the younger sisters with lesser dowries.

In page 19, where the table is, it seems that the last digit got dropped in the Eldest daughter of a rich knight results (rows 4 and 5). The dowries should be £2d6+13 and £2d6+23 treasure, respectively. 2d6+1 & 2d6+2 (average £8 & £9) would mean that the dowry would be on average less than a vassal knight's eldest daughter (1d6+6 -> £9.5 on average), which was not the intention.