However the comparative method is not without flaws, or criticism: Ironically these issues fall into the camp of the comparative method being too similar to genetics, and issues caused by the comparative method not being similar enough.
In the former camp, many modern day linguists consider the tree model implied by the comparative method as overly simplistic, and express doubts that genealogical comparison like is misleading. A tree model implies a series of distinct nodes: It implies sudden and irrevocable change in a population preventing further contact ie many different proto-languages coexisting without further interaction, which is clearly misleading in areas of continuous landmass. Critics claim the model ignores situations where many dialects within a language evolve into distinct languages, over a long period of time where innovations are shared (known as areal diffusion). Populations bordering others have a higher borrowing of linguistic features of each other, as well as similar cultures - influencing linguistic change in a shared direction, this is known as linkage. Different models such as the wave model, which attempts to describe how language features spread over a continuous territory, and are gaining popularity with modern linguists as a way of modelling linguistic linkage and areal diffusion. A good analogy is of a pebble thrown into a pond, the effect is strongest near the epicentre, and weakens as it radially spreads. 
Another problem picturing languages as distinct nodes is the idea of linguistic uniformity within a proto-language, despite the existence of dialects within even small language communities (However the real life implications of this simplification are doubted) \cite{campbell2004}.  
The other main issues with the comparative method come from an underlying assumption that sounds will evolve in a fixed way; despite the reality that linguistic change is far more random and shares a lot of issues with genetics. A good example being analogy, where words of separate and unrelated languages coincidentally converge implying false relatedness. Words and linguistic features are often borrowed from other languages, which can imply a closer degree of relatedness that is correct.