This is of course difficult to prove, but I would say traditional grammars tend to be prescriptive, i. e. they are conceived as a list of rules to be followed in order to speak the language correctly, and usually to establish a standard; deviations from the standard will be either pointed out and discouraged as "mistakes" or "deformations", or simply ignored. (This prescriptivism is the strategy adopted by certain national language academies, like the Academie Française and the Real Academia Española.)
Whereas "scientific" grammars will probably tend towards a proper descriptive approach, a non-judgmental exposition of how the language is actually spoken in its variety of registers: formal, familiar, intimate, slang, etc.; this variety will be acknowledged and studied as part of the language, as will the variety of regional accents, of lexical forms, and the like.
As an example of prescriptive grammar in the face of obvious deviations from the supposed standard, I might offer the fact that traditional grammars in Argentina, where I was born, taught us only the pronouns and verb forms employed in European Spanish, even though my dialect (spoken by maybe 30 million people!) uses different second person pronouns and a very distinctive verb conjugation for them. This was thirty years ago; a few decades before not only was the European standard taught exclusively, but the use of our dialectal forms was pronounced a "linguistic vice" and the teachers were told to erradicate it.
It makes sense, of course, to design grammars that only focus on the standard, especially for second-language students who won't have the opportunity to take in the language naturally and don't have the time to learn all registers. But they shouldn't be in the same category as real scientific grammars.