Because computations with BigDecimal are much slower than double, in terms of memory used and processing time. With appropriate rounding at appropriate times, the small precision loss is negligible (for us).
Yes we use column definition, but our model is huge, the most Studio can help, the most time we save. Additionnaly, you can currently define a @Digits rule that is inconsistent with the specific DDL. At least Studio could rise a warning in that case, but ideally one would define the precision and the type wanted, then Studio would generate the @Digits rule and DDL fields, that could be overriden.