Yepp, at breakfest I realized old truth that energy is not I*t but I*U*t :} - thus the area under coreesponding wave splashes must not obligeously be equal by cm2.
Thus Vroutek was right (respect!) and I must screw on radiator two diodes instead of three - but then max load must be under 27A or solder there 4 diodes and then load must be under 54A (they has specified 37A max).
.
So so... I was expecting to solder three diodes (dsei30) by the beginning, seems I need to order one additional.
And what about website, ought it be corrected if both diodes see equal about 137% of load current??
However the bottom one has longer current and upper shorter, therefore probably, on rights of repetitive surge formula in datasheets, the original website statement may stay as it was. Just not a two times, but slightly bottom-one has more heavier work.
And measurement - I made it on 50Hz, on sinus, and who knows how reacts those soon-after WW2 produced heads, probably there was some resonance effects into the frame or who-know-what else. I believe simulation more than measurehead.
Do now You agree??