Yes that is true, but they would have access to the total data...

Yes that is true, but they would have access to the total data on responses and side effects. The blinding should just be to which arm a patient is in, not the patients response. While not perfect, the total data can give you an idea if the treatment is working. For example, lets say the expected response in the placebo arm is 20 and the average response in the whole study is 10 (i.e both arms at a 1:1 allocation), then it implies the treatment arm average must be 0 (20 + 0 / 2 = 10).

There is some complexity here as sometimes the expected response in the control arm is not known accurately before the study starts. If the expected placebo arm average is 20, but the real value is 40, then looking at the totals can be misleading.

One thing DXB could have done is designed this trial as a Phase 2b/3 where there was a full readout after the first 72 patients followed by recruitment of the remaining patients required for the pivotal study. Given the small number of patients in the original Phase 2 this would have made some sense, the only downside is you need a few more patients (about 10%) to make the stats work.

This is no criticism of what DXB have done. Trial design is a hard problem and given the constraints DXB operates under they have done a good job with the trial to date.