Combining references without the Frankenstein effect

Last week I was sketching a street scene and pulled from three different photo references at once. One had this beautiful converging perspective on the buildings, another had a figure grouping I liked, and the third had interesting light on the pavement. The perspective and figures clicked together almost immediately, but the moment I tried to drop in that lighting from the third photo everything started to feel stitched together instead of whole.

It got me thinking about what actually makes multiple references gel versus fight each other. My instinct says it comes down to keeping the construction logic consistent. If the light direction, eye level, and scale all agree, you can borrow pretty freely. But the second one of those fundamentals contradicts the others, the whole drawing starts looking like a collage instead of a scene.

I am still figuring this out though. When you're pulling visual elements from more than one source, how do you decide what belongs and what to toss? Is there a checkpoint you run through, or is it more of a feeling once you have enough mileage?