H1 ASC-AS_A refined time looks for low threshold crossing rather than high threshold crossing

There was a change made in !115 (diffs) so that now the H1 ASC-AS_A refined time looks for threshold crossing below the mean rather than above the mean (what we actully trigger a lockloss on). In reality, some locklosses go below first but the majority go above first.

Really the best thing to do here is refine using a threshold below and above the mean. This would be a rework on MR !98