At first, I was completely focused on the filter's passband shape.

I measured the ripple, adjusted component values, and finally made the response flat. I thought I had solved the problem.

What I did not study was the absolute insertion loss.

I forgot the most basic thing: a filter is not just about shape. It is about how much signal it lets through.

Then an Italian hobbyist asked me a simple question:

"Why not use air-core coils?"

I had no good answer. And worse – I realized I had never even considered the possibility that the PCB spiral coils themselves were the problem. I was too busy optimizing the wrong design choice.

If he had not asked that question, I might still be tweaking the PCB filter today, never noticing the 20 dB loss hiding in plain sight.

Lesson learned:
Before you start "improving" something, ask yourself whether the thing you are improving should be there at all. Question the design choice itself. Do not assume that because it exists, it is right.

Sometimes the most useful engineering lesson comes from discovering that an optimization was applied to the wrong starting point.