Again, the manual entry describes that:
colorRgb ... An array with 3 elements specifying the red, green and blue color filter of the instance, with color values as floats in the 0-1 range.
Using floats in the range 0-1 is how GPUs actually process the data. It makes many color calculations work more easily, and works regardless of the actual color data being processed. For example when using colors in the 0-1 range, the multiply effect is literally just "a x b". If you use colors in the 0-255 range, that is very much specific to 8-bit unsigned storage - e.g. 10-bit storage would use a 0-1024 range instead - and doing "a x b" will process an entirely different type of effect that isn't really useful - to get a multiply effect with the result also in the 0-255 you'd actually have to do "round(((a / 255) x (b / 255)) x 255)".
So basically everything in computer graphics uses colors as floats in the 0-1 range as it works a lot better, and so does Construct.