For the Space Shuttle, the local acceleration due to gravity was calculated onboard based on the vehicle position. So, as you say, they kinda did "carry a 'Gravitational Ephemeris'”

Because the vehicle's location changes with respect to the Earth as it orbits, the state vector is constantly changing, and nav is continually having to recompute it. This is done in the General Purpose Computers (GPCs) via an algorithm known as “super-g navigation.” The Super-g Algorithm performs the following functions:

a. Given the state vector and gravitational acceleration from the last cycle, a new position vector is estimated using either modeled drag acceleration or IMU-sensed acceleration.

b. The gravity at the new position is calculated.

c. Using the change in gravity from the past to the current cycle, the position and velocity vectors are recomputed.

From Guidance and Control / Insertion, Orbit, Deorbit Training Manual paragraph 2.3

More information about the gravity model is available in the FDO On-orbit Console Handbook paragraph 3.5.1.1

The Orbiter's gravity potential model utilizes an infinite series expansion of Legendre polynomials called Pines method. This recursive algorithm accesses a database of harmonic coefficients arranged in lower-triangular-matrix form. An element of the database is termed Jn,m, where the row n is the harmonic's degree and the column m is its order. When m=0, the coefficient's effect is symmetric about a parallel of latitude, and it's termed a zonal harmonic. When n = m for a matrix diagonal coefficient, its influence is symmetric about a great circle of longitude, and it’s termed a sectorial harmonic. No latitude or longitude symmetry is associated with the action of coefficients having n ≠ m, and they are termed tesseral harmonics. The onboard model is configured with a Goddard Earth Model 9 (GEM9) database truncated to fourth degree and order.

Partial answer:

For the Space Shuttle (a small US fleet of crewed LEO spaceplanes which flew intermittently from 1981-2011), the local acceleration due to gravity was calculated onboard based on the vehicle position. So, as you say, they kinda did "carry a 'Gravitational Ephemeris'”

Because the vehicle's location changes with respect to the Earth as it orbits, the state vector is constantly changing, and nav is continually having to recompute it. This is done in the General Purpose Computers (GPCs) via an algorithm known as “super-g navigation.” The Super-g Algorithm performs the following functions:

a. Given the state vector and gravitational acceleration from the last cycle, a new position vector is estimated using either modeled drag acceleration or IMU-sensed acceleration.

b. The gravity at the new position is calculated.

c. Using the change in gravity from the past to the current cycle, the position and velocity vectors are recomputed.

From Guidance and Control / Insertion, Orbit, Deorbit Training Manual paragraph 2.3

More information about the gravity model is available in the FDO On-orbit Console Handbook paragraph 3.5.1.1

The Orbiter's gravity potential model utilizes an infinite series expansion of Legendre polynomials called Pines method. This recursive algorithm accesses a database of harmonic coefficients arranged in lower-triangular-matrix form. An element of the database is termed Jn,m, where the row n is the harmonic's degree and the column m is its order. When m=0, the coefficient's effect is symmetric about a parallel of latitude, and it's termed a zonal harmonic. When n = m for a matrix diagonal coefficient, its influence is symmetric about a great circle of longitude, and it’s termed a sectorial harmonic. No latitude or longitude symmetry is associated with the action of coefficients having n ≠ m, and they are termed tesseral harmonics. The onboard model is configured with a Goddard Earth Model 9 (GEM9) database truncated to fourth degree and order.

For the Space Shuttle, the local acceleration due to gravity was calculated onboard based on the vehicle position. So, as you say, they kinda did "carry a “Gravitational Ephemeris”

Because the vehicle's location changes with respect to the Earth as it orbits, the state vector is constantly changing, and nav is continually having to recompute it. This is done in the General Purpose Computers (GPCs) via an algorithm known as “super-g navigation.” The Super-g Algorithm performs the following functions:

a. Given the state vector and gravitational acceleration from the last cycle, a new position vector is estimated using either modeled drag acceleration or IMU-sensed acceleration.

b. The gravity at the new position is calculated.

c. Using the change in gravity from the past to the current cycle, the position and velocity vectors are recomputed.

From Guidance and Control / Insertion, Orbit, Deorbit Training Manual paragraph 2.3

More information about the gravity model is available in the FDO On-orbit Console Handbook paragraph 3.5.1.1

The Orbiter's gravity potential model utilizes an infinite series expansion of Legendre polynomials called Pines method. This recursive algorithm accesses a database of harmonic coefficients arranged in lower-triangular-matrix form. An element of the database is termed Jn,m, where the row n is the harmonic's degree and the column m is its order. When m=0, the coefficient's effect is symmetric about a parallel of latitude, and it's termed a zonal harmonic. When n = m for a matrix diagonal coefficient, its influence is symmetric about a great circle of longitude, and it’s termed a sectorial harmonic. No latitude or longitude symmetry is associated with the action of coefficients having n ≠ m, and they are termed tesseral harmonics. The onboard model is configured with a Goddard Earth Model 9 (GEM9) database truncated to fourth degree and order.

For the Space Shuttle, the local acceleration due to gravity was calculated onboard based on the vehicle position. So, as you say, they kinda did "carry a 'Gravitational Ephemeris'”

Because the vehicle's location changes with respect to the Earth as it orbits, the state vector is constantly changing, and nav is continually having to recompute it. This is done in the General Purpose Computers (GPCs) via an algorithm known as “super-g navigation.” The Super-g Algorithm performs the following functions:

a. Given the state vector and gravitational acceleration from the last cycle, a new position vector is estimated using either modeled drag acceleration or IMU-sensed acceleration.

b. The gravity at the new position is calculated.

c. Using the change in gravity from the past to the current cycle, the position and velocity vectors are recomputed.

From Guidance and Control / Insertion, Orbit, Deorbit Training Manual paragraph 2.3

More information about the gravity model is available in the FDO On-orbit Console Handbook paragraph 3.5.1.1

The Orbiter's gravity potential model utilizes an infinite series expansion of Legendre polynomials called Pines method. This recursive algorithm accesses a database of harmonic coefficients arranged in lower-triangular-matrix form. An element of the database is termed Jn,m, where the row n is the harmonic's degree and the column m is its order. When m=0, the coefficient's effect is symmetric about a parallel of latitude, and it's termed a zonal harmonic. When n = m for a matrix diagonal coefficient, its influence is symmetric about a great circle of longitude, and it’s termed a sectorial harmonic. No latitude or longitude symmetry is associated with the action of coefficients having n ≠ m, and they are termed tesseral harmonics. The onboard model is configured with a Goddard Earth Model 9 (GEM9) database truncated to fourth degree and order.