You do wonder if about PAC's logic. Would they rather actual costs had been increased to match the bids? Perhaps they could have gold plated the cabinets. Or employed a large number of consultants at extortionate rates to project manage the whole thing.
Also, it's premature to judge savings at this stage. The cost model was built on an even roll-out of the mix of solutions and environments. With the more cost effective stuff being done first, it seems inevitable that some costs will increase although I still thing the great majority of
Also, as NAC noted, as BT are liable for any downside, that is actuals exceeding what was anticipated, they had to build in a contingency (5-8%). As payments are on actuals, they will only get that contingency if it's necessary. In addition, this isn't calculated nationally (so an overspend in one area would counteract an underspend in another). It's project by project.
Anybody who has been involved in complex bids will tell you they are a great deal of work, and there were dozens in BDUK, all of which will have involved separate assessments according to the local conditions, requirements and local priorities. That's against the background of a lot of unknowns with regard to detail (which is why a model will have been used). It's hardly surprising that there's significant difference between the actuals and bid price and the latter is somewhat higher than the former.