OK, looking at your pictures and as you replace the U4 RAM (considered as safe), the issue should be in the U13 area.
On the pictures, you can notice that 1 pixel on 8 in permanently lighted.
The first 7 pixels are ok, and the last 8th is always lighted.
On the video board, the datas are addressed by the 6845 CRTC from the memory (this chip don't care about the data itself, and is used only for addressing the video RAM). Basically, the datas are serialized/clocked and sent to the display using the RDATA/RCLK signals. The data to be transmitted to the display are selected by U13. This chip is a "data multiplexer" and select one bit among eight bits. The address is given by MA0..MA2 and the source is D0..D7 from the memory:
The "Y" output reflect one of the eight entries (D0..D7) selected by the 3 bits selectors (S0..S2, aka MA0..MA2). Note also that, there is a pullup resistor (R7), so if an input is not wired, the default value at Y is "1"... so a lighted pixel. But on your board, R7 is not mounted (this resistor is not mandatory).
As there is one pixel permanently lighted, the issue is that at a time, the "data bit" is not transmitted but set to "1". As the first 7 bits are OK, and the last 8th is always 1, the culprit line should be D7 (or D0 - I'm not sure of the direction !). This may come from:
- a deffective RAM, but as you change U4, this option should be rejected
- a deffective U13, with a burned entry (D0 or D7
- a broken line at U13 inputs (D0..D7)
The problem could also be from the selection entries (S0..S2), either from deffective U13 or a bad adressing from the 6845, but I doubt for this option, because the ouput will be not always to "1".
So, I'm quite sure that the issue is on the D0/D7 lines, at U13 inputs.
Check with an ohmmeter, the continuity of the D0 (and D7) lines :
- check between pin 1 of U13 and pin 11 of U4 (D0)
- check between pin 8 of U13 and pin 19 of U4 (D7)
If none of thoses lines are broken, this mean that the culprit is the U13 itself, and thus, replace the 74HCT354.