Those numbers are consistent with the specified 12 db of attenuation, and the measurements between the input pins and the ground sleeves of the output connectors correspond, as can be expected, to the sum of the other two measurements. The actual attenuation, btw, will be a bit greater than 12 db depending on the input impedance of the component into which the attenuators are connected, because that input impedance will be in parallel with the 2.19K/2.18K resistors, resulting in a combined impedance that is at least slightly lower than those values.
I'm puzzled, though, by the scale factor business. It shouldn't matter whether you use x10K or x20K, because the numeric reading would change accordingly. But multiplying each of the measured results by the x20K setting you used gives resistances that are much higher than I would expect, and higher than would seem to be desirable, even though they are in the right proportion to each other. If you indicate the make and model of the specific multimeter, it may allow me to shed some further light on this.
Also, I'll mention that when measuring relatively high resistances it would be a good idea to make a point of not having your fingers on the conductive part of at least one of the two multimeter leads. Otherwise the reading may be affected by the resistance of your skin and body.
Best regards,
-- Al