Total radiant flux measurement of an LED with an optometer

[vst98]

Measuring the total radiant flux with an optometer of SFH 4550 LED from osram I'm getting ~70mW,
If I place 5mm long rubber hose around the led (top 2,5mm of the led is free), the power is reduced to 50-40mW.
This does not make sens to me, the led has a half angle of ±3° and emits nothing to its sides,
why is the power then reduced so much ?

 

[CWatters]

Is it getting a lot hotter?

 

[vst98]

I have a constant 100mA current going through the led, that could be a reason, but I have not noticed that.
If a place this rubber hose at the top of the led, so that most of the led housing is not covered, I am again
having significant power loss. I have also placed the led in a small aluminum tube (~10mm long same diameter as the led) and
again there is a power loss. I do not understand, this all should not make a difference if an led emits almost all of its power in ±3°.

One more thing, I have also measured total power flux of VSMY2850RG, a ±10° led. If I place the led at the
entrance of the integrating sphere I have ~30-35mW, only when I place the led completely inside the sphere
the power is ~55mW as stated in the specs. Ok, I think it's now clear to me why, the chip is encapsulated in transparent epoxy, but ...
I wonder why is vishay producing reverse versions of this thing ?
Why it is so hard for led manufacturers to state to which power the intensity plot is related ?