I just wanted to say:
Getting certain attenuation value is relatively easy but getting flatness across wide range is half art, half science. HP attenuators for example, uses copper "flaps" to purposely generate stray capacitance and they are hand tuned. Also, registers like that aren't inductance free. Often carbon/metal coating on ceramic core is cut in spiral to establish values.
Also, OP's construction has way too much stray inductance and capacitance. I don't think suspending resisters in air is bad thing but choice of connectors are not optimum. If BNC, center conductor attach point should be much shorter, and I don't mean resister leads. Internal to that connector is a long enough lead wire between gold plated brass center pin and where it comes out of the metal shell.
Argh.... I should stop reading every post on EEVBLOG! Now I want to replicate this experiment myself and see how I do!