In picure rise time 20% - 80% with Tektronix TDS3032. Tester MK2 is connected to scope with 20 cm RG174 cable and 20 dB attenautor.

That seems in good agreement with the nominal Bandwidth of 300MHz.

For tau rise from 20% to 80% and 300 MHz BW the rise time should be

(ln(0.8/0.2)/(2.pi))/300MHz = 735 psecs and you're measuring around 700 psecs which seems pretty close.

Don't you need to subtract out the calibrated rise time of the pulser from the displayed rise time to get the real rise time first?

I'm not an expert on this but I think it is not a simple subtraction. For uncoupled systems I think it is the square root of the sum of the squares of the separate taus.

So, if that is the case, you need to square the tau, subtract the square of the pulse generator tau and then take the square root.

see the bit on cascaded blocks in Wikipedia:

http://en.wikipedia.org/wiki/Rise_timeIf the pulse generator tau is around 200 psecs then 700 psecs should in fact be 671 psecs. The 960 psec should be 939 psecs.

So, yes you're correct in that they need correcting but the difference is not huge. It would be much more significant for higher BW scopes.

Doing these corrections the BW of the nominal 300MHz scope is around 330MHz and the BW of the 200MHz is 235MHz which seem very reasonable numbers given that I guess manufacturers aim to be plus 10% to ensure that they meet the specification and also it is a pretty approximate measurement in the first place and is making the approximation that the behaviour is governed by a single pole (is essentially a simple RC circuit).

I'm quite impressed by how close it gets.