Hi Vlasios,
you're right: though I don't know the details how the manufacturers exactly measure the
PMT QE, the information I have is that they indeed illuminate the PMT "head-on".
However, I guess they characterize the photocathode material, not really each PMT. Hence
I suppose they use for practical resons small PMTs (1-2 inches), not the emispherical
ones. In this case the interval of angles of incidence is limited. However I might be
wrong! This is indeed an information to check.
Anyway, there is another strong argument why I assume that the measured QE applies for
"normal" or "near-normal" incidence: the manufacturers illuminate the PMT in air. Now,
the optical model of the photocathode shows that for a PMT operated in air the response
is a very flat function of the angle. Normal incidence, or -say- 30 degrees makes a very
little difference. This is true for a PMT coupled to a liquid as well, as far as the
angle of incidence is lower than the critical angle that arises from the interface
liquid-vacuum. But for higher angles the response varies a lot. It is especially this
strong variation that the model tries to reproduce.
Summarizing: the manufacturer measures the QE of a photocathode by using (probably!)
small tubes and thus illuminating at "near-normal" incidence. Furthermore, under normal
experimental conditions it can be shown that the photocathode response is fairly
constant over a large interval around the normal incidence (up to ~ 40 degrees).
Please, note he discussion here neglects any effect due to different recollection
efficiency for the photons transmitted inside the PMT (this is too difficult to handle!).
Since you show a great interest for all topics concerning the PMT modeling, I dare to
advise you giving a look at my recent NIM paper on this subject:

D. Motta and S. Schönert, "Optical properties of bialkali photocathodes", Nucl. Instr.
Meth. A, 539 (2005), 217-235.

You can also find there some references to other interesting papers on the topic.
Have fun!
Dario

> Hi Dario,
> 
> First of all thanks for your detailed answer.
> 
> I could verify that all the CE>0.5 events were from photons coming from the inside. 
> 
> I have one more question though: why do you assume that the QE is given for normal 
> incidence? I thought that the manufacturer is measuring the QE by illuminating the 
> *whole* face of the PMT with a parallel beam of light. So the measured QE comes 
> from many different incidence angles. 
> 
> Vlasios