`big_value` `comparison` silently reports 0.00% when the period comes from a `range_calendar` filter

Summary

comparison={ compare_vs="prior period" } works when the period is set by an explicit date_range, but reports a flat 0.00% when the same period comes from a range_calendar filter via filters=[...]. The value itself is correct in both cases — only the comparison is wrong, and it fails silently rather than erroring or omitting the delta.

Environment

• Evidence CLI v0.9.3 (latest)
• BigQuery direct connector

Steps to reproduce

{% range_calendar id="dates" value_column="start_date" default_range="last 7 days" /%}

<!-- A: period from the filter -->
{% big_value data="v" value="sum(amount)" fmt="usd2"
    filters=["dates"]
    comparison={ compare_vs="prior period" delta=true } /%}

<!-- B: period from date_range -->
{% big_value data="v" value="sum(amount)" fmt="usd2"
    date_range={ range="last 7 days" date="start_date" }
    comparison={ compare_vs="prior period" delta=true } /%}

Actual

Both render the same value ($6,573.38), confirming the filter applies correctly, but:

• A → ▲ 0.00% vs. last period
• B → ▲ 4.04% vs. last period

B is correct.

Expected

A should resolve the prior period from the filter's active range and report 4.04% as well. Failing that, it should omit the comparison rather than render a confident, wrong 0.00%.

Impact

Medium-high. This is the natural way to build a dashboard — one date picker driving every tile — and it silently reports "no change" on every KPI. A reader has no way to tell the delta is broken. In a dashboard with a shared date filter, comparison is effectively unusable, and its failure is indistinguishable from genuinely flat metrics.