I agree with Marc's points 1, 2, and 4.

For point 3, SRFI 26 leaves <> unbound, whereas this SRFI should bind _.  We can go with _ and bind it to a syntax-error, but if we want to use <> we need to update SRFI 26.

For point 5, I don't think grabbing a non-alphabetic symbol is a good idea.  This SRFI is about predicates, and I think it's better to limit it to just predicates.

Point 6 is really out of scope.  The rationale for SRFIs is to put forth ideas that people may wish to use.  The down side of extensible syntax is that if you don't know what the macro means, you can't understand code that uses it, but that's a price we (usually) gladly pay in Scheme.  It's too much to ask a SRFI author to predict whether a new feature will become popular.

I would add two more points:

7) I think that having (is foo) imply (is equal? foo) is just a mistake.  Equal? is a partly implementation-defined predicate; if you want it, you should specify it.  So `is` with a single argument should just be an error.

8) It seems clear to me that (is _ related-to? _) should simply mean `related-to?`.  When _ is used as a place holder, it is always assumed that it might mean different things in different places.  The interpretation (lambda (_) (related-to? _ _)) violates this assumption.