IC-9700 + vhfdesign LNA 23cm EME 1st SAW measurement

I was wondering if LNA mounted 50 cm from the IC-9700 + 15m Ecoflex 10 makes sense... Tests using setup like this (see picture)

shown it makes sense... well how was I testing it? I've installed a 23 cm beacon about 70 km from my home qth. Beacon has power lower than 100mW... and with LNA I see it on the IC-9700 waterfall and can hear... while without it...rarely.

So I decided to make a more LAB test;) This is my setup:

Signal generator HP 8642B frequency 1296200 kHz, amplitude - 140dBm.

About 2 meters of RG-223 cable.

LNA (vhfdesign 23cm EME 1st SAW - 18dB gain 0.3 NF) with relays that you can see on picture above (with some semi-rigid).

About 1 meter of RG-223 cable.

IC-9700 run on 1296200 kHz in CW with 500Hz filter and AGC-OFF.

USB cable and computer with Linux and baudline (audio spectrum analyzer).

Results... (distance from top of background to maximum peak)

LNA only -  22.14 dB, background ~ -56 dB

PREAMP only - 18.56 dB, background ~ -65 dB

no LNA no PREAMP - 13.86 dB, background ~ -65 dB

Pictures:

Conclusion? For me very simple, it confirms what I see on my antenna and weak beacon signal, LNA even located 50 cm from IC-9700 is making a noticeable difference BUT.... as noticed by Staszek SP6GWB (thx!) strange thing is that when using LNA background noise is about 10dB higher than without... 

So I was forced to do another test. Just LNA, no relays, not bias-t. Please take a look at the drawing below.

And new measurement also below... unfortunately nothing changes ;( (this was taken in two approches, using power over built-int bias-t of LNA and using external power - but no difference)

LNA NO-RELAYS NO BIAS-T -  22.23 dB, background ~ -54 dB

And one more test... 

Instead of signal generator (HP 8642B) I've connected dummy load. Unfortunately the same result. So ... this ~10 dB increase of noise is coming directly from LNA.

If i leave LNA INPUT unloaded - no generator, no dummy load, then background goes down to ~62 dB.

And here is a movie from measurement of RF noise level... ok, the Spectrum Analyzer is old but analog and it really can be seen that if LNA is terminated with 50 ohm it is inserting its own noise of level of at least 5 dB.... and most probably this is around 10 dB but cant be seen here due to dynamic range limitation of spectrum analyzer. Many thanks goes to Mariusz SP6RGV for support with measurements.

Conclusion again? None, whats goin' on here? Is it improperly matched input? For sure the problem is not related to the radio.

OTOH is it really a problem? I do not know, I asked vhfdesign for comment.

Summary table.

Input signal	LNA or IC-9700 PREAMP	Signal to Noise distance	Background noise
-140 dBm	LNA only	22.14 dB	-56 dB
-140 dBm	PREAMP only	18.56 dB	-65 dB
-140 dBm	LNA/PREAMP off	13.86 dB	-65 dB
-140 dBm	LNA without relays and biast-t	22.23 dB	-55 dB
50 ohm dummy load	LNA without relays and biast-t	n/a	-55 dB
input open	LNA without relays and biast-t	n/a	-62 dB