The problem is how do we know that all water is the same. Why must the different instances of natural kinds (like water) all have the same meanings. More so, which properties can be identified as the essential properties of any particular natural kind? These questions are critically raised by D. H. Mellor, who throws the ball into the court of the MN's to uphold their theory against RNT's.
Mellor states that their (natural kinds) meaning do not depend on the existence of archetypes. Furthermore, even if they did, archetypes still do not produce enough evidence for the existence of a natural kind's essence (or essential properties.) For Putnam's theory to produce an essence, the similarities among the natural-kinds must be transitive, so that all possible F's share the same property G with each other. The problem is in two natural-kinds that are different, but one having some of the same archetypal properties with the other in the real world.
For further clarification of Mellor's criticism, let's go back to the example of water. We can say that water is this big clear, wobbly thing, that is reflective and gathers in large quantities like oceans and lakes. In another world, there might be a natural-kind (say alcohol) having the same archetypal properties as water on this world. So, in other words, they are not being compared directly, only through their archetypes.