This graph shows that the mask in U1 maxs out around kl0=23.5.
My conclusion is that the ANFP really specifies modem behaviour in the formulae in the table; the graphs are just a few examples made using those formulae.
For upstream, I don't (yet) know how many decimal places kl0 can be calculated to (the formulae in the table runs to 5 significant figures). But we do know that the table copes as it varies between 0 and (just under) 24.
Individually, and slightly more accurately, it looks like U1 changes as kl0 varies over a range of 0-23.6, whereas u2 changes as kl0 varies over a range of 0-8.5.
For downstream, we know that CAL can vary from 0 to 52, but only in integer steps of 2, so we know there are exactly 27 graphs there.