Apologies, I didn't quite answer your question fully.
The difference between my net debt forecast and the reported outcome is due to two major variances:
1. Working capital, the company ending the period with some $8m less of it than I had forecast (essentially, a 40bp difference in Working Cap-to-Sales terms). So, 2H2017 OCF was $116m compared to my $103m forecast.
2. As you rightly deduced, the other major variance was that my capex estimate was for $70m in the second half (it was $49m in 1H2017), while the actual result of $47m was actually less than 1H2017. So there is a further $23m of net debt variance explained, I clearly gauged the timing of the capital expenditure on the new factory incorrectly. But it is really a mere timing issue, because for the current half, I had only $30m of capex in my model, and I now think it will be about $50m. So, really, it's just a 6 month lag of some $20m of capex... largely "forecasting noise".
The significant aspect of what is happening with this company here is testimony to its prolific cash flow generation.
For here is a business that is undertaking a once-a-generational total renewal of its manufacturing base, while continuing to pay out ~75% of its profits to the owners of the business, and yet over the period this is all happening, Net-Debt-to-EBITDA will still have been reduced: