Let me give it a try. It appears that Van der Tak's analysis is not addressed at the Tal-Portisch game (oh, my memory), but Nedeljkovic-Volpert, Belgrade 1961. This is the mistake of Gufeld though. Portisch played ...a6 while Volpert didn't. The improvement comes at move 31 (in the 16.Nd5 move order at move 29)


This is not complete; some lines continue to move 47. Not all lines are from Van der Tak himself; he refers to Averbach and Slutsky/Shereshevsky.
Then Van der Tak gives the game Rittner-Arndlind, corr BdF 1970. Alas it happens not to be in my database and 70 moves of typing is too much for me, sorry.

I own some probably unknown analysis by Dutch theoretician Van der Tak (published in Schaakschakeringen, the magazine of the Dutch corr. chess federation NBC) that suggested that the famous line 14.Rhe1 Bxb3 15.cxb3 b4 of Tal-Portisch, Oberhausen 1961, should lead to a draw.
The problem remains that Black has hardly winning chances.