[Kiril1Shevtsov]

A telescopic overnight coupon with positive lockout can omit the observation immediately before the frozen tail. Its repeated tail fixings can then use a front-stub fixing instead of the last pre-lockout fixing. Retain one additional business-day observation when lockout is positive; the zero-lockout path is unchanged.

The existing overnight-leg NPV test encoded the wrong anchor. Update its reference using an independent daily SOFR deposit product and add an uncompressed-leg control, keeping the absolute tolerance at 1e-8.

Validation against cf74c9f657150750dfbf00ed6fd4c0768bc5ce36:

• Separate clean baseline and candidate shared-library and stock-test builds on Linux x86_64 with GCC 14, C++17 and Boost 1.87 (default configuration, CMake unity build, -O1 -g0 -DNDEBUG, warnings as errors).
• Full candidate stock suite: 1,714 test cases and 619,739 assertions passed, with no failures or skips. The baseline suite also passed; the updated NPV regression fails on the baseline while its uncompressed control passes.
• Independent date-geometry checks: 4,130 checks passed. Independent daily-product and NPV checks: 52 checks passed.

This change addresses the missing pre-lockout anchor. Separate nonzero daily-spread discrepancies remain outside its scope; the broader control matrix still has 80 such failures. No claim is made for aged telescopic coupons or cutoffs inside an omitted middle interval.