The English negative comparative-modified numerals (CMNs) "no more than n" and "not more than n" different in surprising ways: The former conveys exactness (EX) and evaluativity (EVAL), and the latter conveys no exactness (NO-EX) and ignorance (IG). The existing literature offers 3 proposals, one based on Horn (1972), one based a combination of Horn (1972) with the Universal Density of Measurement hypothesis, and one based on a combination of Horn (1972) with the Modifier Alternatives hypothesis. In this paper I argue that both modifications of Horn (1972) are in fact too drastic. And point out that none captures both EX and NO-EX, nor do they make any provisions for IG or EVAL. I propose what I argue is a smaller departure from Horn (1972), and I show it captures all of EX, NO EX, IG, and EVAL. The explanation crucially relies on a combination of Horn (1972) with Negation Alternatives - the idea that for some negative expressions the set of scalar alternatives include variants with the negation as well as without - plus silent exhaustification with O(nly) and E(ven).